Podcast Episode 38

Episode 038 50 min

Nick Sullivan on Cryptography, Cloudflare, and Building a Better Internet

with Nick Sullivan · fmr. Cloudflare

0:00 / 50:00

Also on Spotify, Apple Podcasts, YouTube, Podbean, Player FM

Episode 38 of The Security Podcast of Silicon Valley, published March 9, 2024. Host Jon McLachlan talks with Nick Sullivan, who led cryptography and then research at Cloudflare, about pure math at Waterloo, their years on Apple's DRM team, Cloudflare as a proxy for a big share of the web, CFSSL, TLS 1.3 and Privacy Pass, measuring connection tampering from the server side, iCloud Private Relay, research on a five-year clock, the slogan we make friends, Messaging Layer Security, the Heartbleed Challenge, advising startups, and three prongs of what comes next: post-quantum cryptography, privacy-enhancing technologies and confidential computing.

Episode notes

In this enlightening episode of The Security Podcast of Silicon Valley, a YSecurity.io production with your host Jon McLachlan, we are honored to host Nick Sullivan, a trailblazer in the realm of cryptography and security. Formerly the Head of Cryptography at Cloudflare and now a pivotal figure in startup advisory and angel investing, Nick shares his journey from delving into pure mathematics at Waterloo to shaping the security landscape of the internet. He reflects on his early career, from researching cryptography to his impactful roles at Symantec, Apple, and eventually Cloudflare, where he spent over a decade innovating and advising on next-gen security protocols.

Nick offers an insider's view on the challenges and triumphs of developing Cloudflare's robust security infrastructure, emphasizing the importance of privacy, research, and the collective effort to safeguard the web. He discusses groundbreaking work on global data insights, the evolution of internet security practices, and his advisory role in nurturing startups toward solving today's most pressing security dilemmas.

Join us as Nick Sullivan imparts wisdom on the significance of curiosity, the power of collaboration, and the unyielding pursuit of building a better, more secure Internet. This episode is a must-listen for anyone passionate about the future of technology, privacy, and the endless possibilities that cryptography offers in securing our digital world.

Hosted by Jon McLachlan — co-founder of YSecurity.

The deep dive

We make friends

Nick Sullivan, who led cryptography and then research at Cloudflare, with Jon on pure math at Waterloo, Apple's DRM team, a proxy in the middle of the web, CFSSL and TLS 1.3, Privacy Pass and iCloud Private Relay, connection tampering seen from the server side, a free tier as an immune system, research on a five-year clock, two kinds of security people, cryptography as magic, standards and Messaging Layer Security, Heartbleed and the power of the crowd, advising startups, trust as currency, post-quantum cryptography, privacy-enhancing technologies, confidential computing, and following your curiosity.

Episode 38 of The Security Podcast of Silicon Valley aired on March 9, 2024, with Jon McLachlan hosting. The guest was Nick Sullivan, who spent a decade at Cloudflare as head of security engineering, then head of cryptography, then head of research, and was advising the company and startups when they spoke (Podbean, Apple Podcasts, Spotify). The episode is called “Nick Sullivan on Cryptography, Cloudflare, and Building a Better Internet.” It has no video version. Jon and Nick go back to Apple, where both worked on the DRM team.

The line that sums up the conversation is the slogan Nick gave his research team at Cloudflare.

“one of the slogans that I came up with for the research team was, we make friends.”

Nick Sullivan, at 16:34

Listen · Nick Sullivan on episode 38, from 16:20
About 30 seconds: Nick on why research runs on relationships. The clip stops on its own. Read the full transcript.

This page follows the conversation in order: a section for each part of it, thirteen figures, thirteen clips from the tape, a LinkedIn post, the full transcript, and five tools: a quiz, a scorecard for whether your security makes friends, a calculator for how many handshakes a group chat needs, a timeline, and a to-do list. Where the decodes of the tape disagree on a word, the transcript puts its best guess in brackets, and the page doesn’t quote or retell those words.

Pure math first

Jon opened with a career built on security and cryptography, and asked whether Waterloo had been about cryptography. Not quite:

“I studied pure mathematics, actually.”

Nick Sullivan, at 0:33

“it is much more abstract and much more theory focused.”

Nick Sullivan, at 0:40

Jon had no complaint about that order of operations.

“before you can get into the bit smashing, right?”

Jon McLachlan, at 0:46

Nick’s LinkedIn lists a BMath with distinction from the University of Waterloo in pure mathematics, combinatorics and optimization, and an MSc in computer science and mathematics from the University of Calgary, specializing in cryptography, with a thesis on fast arithmetic on elliptic curves over prime fields (LinkedIn). Jon ran through what came next: cryptography researcher, security analyst at Symantec, then Apple.

A timeline. May 2004: CISaC, cryptography research on elliptic and other algebraic curves. December 2006: Symantec, security analyst, and a co-author of the Internet Security Threat Report. July 2007: Apple, the DRM team, with Jon on the obfuscating compiler team. July 2013, filled: Cloudflare, head of security engineering, with the company under fifty people. December 2015, filled: head of cryptography, reporting to the CTO. August 2019, filled: head of research, he founds Cloudflare Research. March 9, 2024, filled: episode 38, Nick on this show. March 2026, filled: Internet Architecture Board, a two-year term. Then: "I studied pure mathematics, actually." Nick, at 0:33
Nick's path from pure math to the Internet Architecture Board. Sources: his LinkedIn and the episode.

The DRM team

“We were both on the DRM team.”

Jon McLachlan, at 1:08

Nick worked more on the cryptography and the business logic, Jon said, and Jon was on the obfuscating compiler team, which is harder to talk about. They worked closely together. Nick’s LinkedIn gives his Apple title as “Mathemagician,” from 2007 to 2013, and says he built the core DRM technology behind iTunes Movie Rentals and Purchases, iBooks, App Store receipts and iTunes Radio, with custom cryptographic protocols and services that handled over a trillion transactions a day (LinkedIn). FairPlay is Apple’s family of DRM technologies, for videos, books, apps and music (Wikipedia).

“Apple Media Empire is built on the foundations of having secure ways of enforcing business logic.”

Nick Sullivan, at 1:28

Jon noted the strong political opinions on whether DRM is good or evil.

“Maybe it’s both.”

Jon McLachlan, at 1:46

Without getting into the politics, he said, the technical challenges were open and interesting. Then Nick had the good sense to leave for a young, budding Cloudflare. If you ship software that people will try to take apart, our anti-reverse engineering work starts where that team’s did.

A proxy in the middle

“I joined when it was a small company, under 50 people or so.”

Nick Sullivan, at 2:55

“it basically sits as a proxy between people and websites or web resources.”

Nick Sullivan, at 3:03

That generic framework makes it powerful, he said. In the middle, you can do security monitoring, run web application firewalls and stop denial of service. Wikipedia describes Cloudflare’s services as acting primarily as a reverse proxy between website visitors and a customer’s hosting provider. Matthew Prince, Lee Holloway and Michelle Zatlyn founded the company on July 26, 2009, and it went public on the New York Stock Exchange in 2019 (Wikipedia).

The web is under constant attack, he said. Attackers keep finding new loopholes to steal identities and money, and a security blanket over your website and web services helps.

Listen · Nick Sullivan on episode 38, from 2:52
About 35 seconds: Nick on what a proxy in the middle can do. The clip stops on its own. Read the full transcript.

Trust plumbing

His first years were heavy engineering, he said, on the security architecture that was missing.

“So I built Cloudflare’s public key infrastructure”

Nick Sullivan, at 4:05

He also built a small team around an open source tool called CFSSL, for creating certificates and managing internal trust, and it helped a fledgling company take off. His LinkedIn adds Red October, an open source take on the two-man rule, and work on Keyless SSL, which let Cloudflare serve HTTPS without holding a customer’s private key (LinkedIn). On September 29, 2014, Cloudflare made SSL free for every site with Universal SSL. The release said fewer than 0.4 percent of websites were encrypted before the launch, and it almost doubled the encrypted web in a day. Nick, then Cloudflare’s security engineering lead, said the company made “cutting-edge cryptography” free and easy for anyone (Cloudflare). Alan Braithwaite remembered Nick from that same year, on the interview panel that hired him, on episode 50.

Three boxes joined by arrows. People, browsers and apps. Cloudflare, filled, the middle party. Websites, and web resources. Below, a list. Filled: in the middle, security monitoring, web application firewalls, and stopping denial of service. Underneath: Cloudflare's public key infrastructure, and CFSSL for creating certificates and managing internal trust. September 29, 2014: Universal SSL makes SSL free for every site, and fewer than 0.4 percent of websites were encrypted before. Then: "it basically sits as a proxy between people and websites or web resources." Nick, at 3:03
Cloudflare in the middle, and the trust plumbing underneath. Sources: the episode, Wikipedia and Cloudflare.

Bridging the gap

Bringing security, cryptography and privacy to a wide audience as a product was one part, he said. The other was the industry’s problems themselves. He worked with standards bodies, tested new ideas with academic researchers and built prototypes.

Ideas that academia had explored with rigor were not always ready to deploy at scale, and he helped bridge that gap. His own focus was cryptography, which led to the head of cryptography role. In that role, his team

“participated in TLS 1.3, invented stuff like Privacy Pass.”

Nick Sullivan, at 5:35

TLS 1.3 was defined in RFC 8446 in August 2018 (Wikipedia). Privacy Pass started as a 2018 paper by Alex Davidson, Ian Goldberg, Nick Sullivan, George Tankersley and Filippo Valsorda: a protocol that hands users anonymous tokens for solving a challenge, which they can spend later to skip challenges like CAPTCHAs (Cloudflare Research). The IETF published the Privacy Pass architecture as RFC 9576 in 2024 (Cloudflare Research). The model, he said, was long-term research based on

“collaboration across the industry and not competition”

Nick Sullivan, at 5:53

That meant the IETF, conferences and academics, and an internship program with probably 40 interns at least over the last couple of years. The team grew past cryptography and protocols into DNS, network measurement, distributed systems and privacy-enhancing technologies. Taher Elgamal, who helped make SSL a standard, talked about where it came from on episode 61.

Tampering at scale

He still consulted with Cloudflare, he said, and a recent paper from his work there measured tampering: participants on the network stopping traffic from getting through, in various parts of the world. Cloudflare sat in front of almost 20% of the web at that point, he said.

“And it’s a lot higher in certain countries.”

Nick Sullivan, at 7:22

The reasons vary: corporate security that keeps employees off phishing sites, geopolitics, and schools that filter content. Sitting in front of so much of the web gave the team the data to see it.

“Yeah, you have just tons of data, I imagine.”

Jon McLachlan, at 7:56

The paper is “Global, Passive Detection of Connection Tampering,” presented at ACM SIGCOMM 2023, September 10 to 14 in New York, by Ram Sundararaman, Louis-Henri Merino, Kevin Bock, Marwan Fayed, Dave Levin, Nick Sullivan and Luke Valenta. It calls itself the first global study of connection tampering through passive analysis of traffic received at a global CDN (Cloudflare Research).

Seen from the server side

Jon asked how you could tell something was being censored inside a school, if the packets never reach Cloudflare.

“It’s very counterintuitive that one would be able to learn this”

Nick Sullivan, at 8:29

Most measurement until then had been active, he said: groups like OONI put devices in people’s homes, or let people opt in, and test a preset list of names. The Open Observatory of Network Interference has monitored internet censorship since 2012, when it launched under the Tor Project (Wikipedia). The new study needed only the server side of the connection, where it looked for signatures. Nick credited the folks at the University of Maryland with the machine learning analysis of TCP flows.

“There’s no authentication on any of these TCP things.”

Nick Sullivan, at 9:44

A request names the site it wants in the handshake. Server Name Indication is the TLS extension that carries that hostname, and it isn’t encrypted, so a passive eavesdropper can see it. Wikipedia notes that security software has used that for filtering and monitoring, and governments for censorship (Wikipedia). A device in the path matches the name against its list and sends a reset, or does other things to confuse the state machines on either side. The handshake gets through, then a stray reset shuts the TCP connection down. Many of those signatures are high confidence, he said, and correlate with particular products that people sell.

Listen · Nick Sullivan on episode 38, from 9:47
About 30 seconds: Nick on how tampering looks from the server. The clip stops on its own. Read the full transcript.
Four boxes joined by arrows. The handshake names the site. A device, filled, matches the name. A reset, filled, or a stray packet. Shut down, dashed, the TCP connection. Below, a list. Active measurement: devices in people's homes test a preset list of names, as OONI does. Filled: passive measurement, tampering signatures, sequences of packet headers, seen at the server across Cloudflare's traffic. Why: corporate security against phishing, geopolitics, and schools that filter content. Then: "There's no authentication on any of these TCP things." Nick, at 9:44
Connection tampering, as Nick described it, and how the study spots it. Sources: the episode, Cloudflare Research and Wikipedia.

A free tier as an immune system

“I always say that’s one of the superpowers of Cloudflare”

Nick Sullivan, at 11:05

He meant the free customer base. Millions of customers can join and use the service. They get a lot of value, and Cloudflare pays a cost for it, and in return it sees what threats are going on across the internet.

“So it’s kind of like an immune system, if you will.”

Nick Sullivan, at 11:26

A threat hits a free customer, he said, and Cloudflare can study it, build a mitigation, and protect its large enterprise customers.

“It’s a really good symbiotic relationship.”

Nick Sullivan, at 11:37

Listen · Nick Sullivan on episode 38, from 11:05
About 35 seconds: Nick on the free tier as an early warning system. The clip stops on its own. Read the full transcript.

Five years to Private Relay

Some of his most recent work, he said, depended on a company whose business is making the network work the way it should, with no money made from tracking individuals.

“Cloudflare participated in iCloud Private Relay”

Nick Sullivan, at 12:14

He described it as a privacy proxy that uses CDNs and network providers, naming Cloudflare, Akamai and Fastly, to hide the client’s IP address, and it’s native on iOS now. Private Relay is part of an iCloud+ subscription. It sends Safari browsing, DNS and unencrypted app traffic through two relays: an ingress proxy run by Apple sees your IP address but not your destination, and an egress proxy run by a provider other than Apple sees the destination but not your IP. Apple announced it in June 2021 (Wikipedia).

“And that kind of happened through five years from when we came out with Privacy Pass and published it.”

Nick Sullivan, at 12:31

It took a lot of conversations behind the scenes about secure DNS and privacy, he said. Apple and Cloudflare were aligned.

“Neither company is trying to monetize individuals”

Nick Sullivan, at 12:51

Four compartments in a row. Your device, IP and destination. Ingress relay, filled, Apple, sees your IP. Egress relay, filled, another provider. The website, sees a temporary IP. Below, a list. The egress relay sees the destination, not your IP address. Part of an iCloud+ subscription, for Safari browsing, DNS and unencrypted HTTP app traffic, announced in June 2021. Filled: Nick's account, five years from Privacy Pass to Private Relay, with providers he named as Cloudflare, Akamai and Fastly. Then: "Neither company is trying to monetize individuals" Nick, at 12:51
Two relays, two halves of what a provider could learn. Sources: Wikipedia and the episode.

A five-year process

That’s the kind of win a research team should expect, he said, on a timeline an executive team funding it has to be comfortable with.

“This is a five-year process.”

Nick Sullivan, at 13:15

Most of the successful projects at Cloudflare took “three to five years to become fully robust,” he said.

Jon asked whether that was a hard sell to the executives. The mission helped, Nick said:

“Cloudflare’s mission statement is to help build a better internet.”

Nick Sullivan, at 13:45

Better can mean more secure, more available, or more private. Research teams are a cost center, he said, and a company growing as fast as Cloudflare did for ten years has a million tensions.

Funding, priorities and communication look different at 100, 500 and 2,000 people, he said. Long-term research is a big swing that few companies are comfortable taking, given the pressure on revenue, cash flow and growth.

“But I think we showed that research can pay off, even in a smaller company, even in a growing company”

Nick Sullivan, at 15:36

He took lessons from research groups at places like Google and Microsoft, and from the Bell Labs days. Bell Labs is credited with the transistor, the laser, information theory and Unix, among much else (Wikipedia).

Listen · Nick Sullivan on episode 38, from 13:01
About 20 seconds: Nick on the clock a research team runs on. The clip stops on its own. Read the full transcript.
Four boxes joined by arrows. An idea, explored in academia. A prototype, built with academics. A standard, filled, through the IETF. A product, filled, in wide use. Below, a list. Nick: most successful projects at Cloudflare took three to five years to mature. The mission: help build a better internet, which can mean more secure, more available or more private. Filled: the cost, research is a cost center, and it runs on relationships. Then: "This is a five-year process." Nick, at 13:15
Research on a five-year clock. Sources: the episode, Cloudflare Research and Wikipedia.

We make friends

Without short-term deadlines, he said, research can’t promise deliverables within even six months, so it leans on relationships and communication inside the company. People need to know the value it brings and want to work with you. Hence the slogan.

“And there’s absolutely no way to do some of the more difficult, long-term things that take away from other priorities that come with research”

Nick Sullivan, at 16:40

Those take people and their support, he said.

Two kinds of security people

Jon liked the slogan, and told a story he likes to tell. One kind of security person walks into a meeting room and everyone thinks, oh no: new standards to implement, a deadline pushed back by a couple of months.

“it’s just going to be roadblock after roadblock.”

Jon McLachlan, at 17:28

The other kind walks into the same meeting with exactly the same expectations for the product, and the room is glad to see them. Jon used Nick as the example: someone who helps the team build the new thing in a way that fits its security goals and its customers’ expectations.

“And Nick is the guy who’s going to help us understand and navigate that complex space as an ally.”

Jon McLachlan, at 18:01

“I’d like to think that I try to embody that second person”

Nick Sullivan, at 18:07

The scorecard on episode 35 asks whether your security enables or restricts, and the scorecard below asks whether yours makes friends.

Listen · Jon McLachlan on episode 38, from 17:30
About 35 seconds: Jon on the security person people are glad to see. The clip stops on its own. Read the full transcript.
Two cards side by side. The first one: the room expects new standards to implement, a deadline pushed back by a couple of months, and roadblock after roadblock. The second one, filled: same expectations for the product, and the room is glad, because this person helps them build the new thing to fit security goals and customers' expectations. Below: Nick, a security person who isn't a resource for the company is potentially a burden. Then: "And Nick is the guy who's going to help us understand and navigate that complex space as an ally." Jon, at 18:01
Two security people walk into the same meeting. Source: the episode.

A resource or a burden

If people come into the meeting expecting trouble from you, Nick said, you won’t get far in a company.

“So you’ve lost already, like, at that point.”

Jon McLachlan, at 18:35

His approach is collaborative, inclusive and helpful, and he encourages his team to help anyone who asks. Research and security groups hold specialized knowledge that a typical engineer won’t have, and a short talk with an expert can save a team from easy mistakes.

If you aren’t a resource for your company, he said:

“you’re potentially a burden, right?”

Nick Sullivan, at 19:32

That’s worse when your group’s goals look like they run against the company’s, he said. Jon called it a dangerous place to be, especially in a contracting market with layoffs going around.

“efficiency is super important.”

Nick Sullivan, at 20:16

Everybody wants a team that can do more with less, he said.

A series of doors

Jon asked whether a moment tipped him from mathematics into security.

“It was kind of a series of events and doors that opened up.”

Nick Sullivan, at 20:50

In his last semester at Waterloo, after pure math and combinatorics, all pen-and-paper abstract work, he took the intro cryptography class from Alfred Menezes, one of the cryptography professors there. Menezes, he said, invited him back to give a guest lecture a couple of years before the episode. Menezes is a professor in Waterloo’s Department of Combinatorics and Optimization and a co-author of the Handbook of Applied Cryptography (Wikipedia).

“everything just clicked for me”

Nick Sullivan, at 21:18

He could take his hard-learned intuition about prime numbers and discrete mathematics and put it into a concrete application.

Like magic

Cryptography is like magic to him, Nick said. It lets you do digitally what could only be done physically, and the intuition is weird.

“you can take data and make it so that no one in the world can see it.”

Nick Sullivan, at 21:54

Physically, you’d put it in an unbreakable safe. Digitally, math does it. Even basic tools like digital signatures enable capabilities that shape how communication systems are designed.

Listen · Nick Sullivan on episode 38, from 21:40
About 40 seconds: Nick on why cryptography feels like magic. The clip stops on its own. Read the full transcript.
Two cards side by side. Physically: put the data in an unbreakable safe. Digitally, filled: math does it, and digital signatures add more. Below, a list. Cryptography is like magic to him: it does digitally what only physical things could do. Filled: even basic tools like digital signatures shape the architecture of communication systems. Then: "you can take data and make it so that no one in the world can see it." Nick, at 21:54
What cryptography does digitally, and its physical stand-ins. Source: the episode.

How software really works

After that, he did a master’s degree in cryptography and moved from math into computer science. A few courses got him into antivirus, exploits and security.

“there’s a way that you think software works, and then there’s a way that software really works.”

Nick Sullivan, at 22:41

He finished the master’s in elliptic curve cryptography and started on implementation, then on threats and the internet. Elliptic curves were suggested for cryptography independently by Neal Koblitz and Victor Miller in 1985, and give equivalent security with smaller keys than RSA (Wikipedia). Symantec hired him as a security analyst and one of the co-authors of its annual Internet Security Threat Report. His LinkedIn lists volumes XI and XII (LinkedIn).

“this kind of opened the door to, like, massive data problems, and data as a tool for explaining the world of security.”

Nick Sullivan, at 23:10

Cryptography solves problems with magical primitives, he said. Meanwhile most of the world lives and builds online, and nothing was built perfectly.

“Everybody’s taking shortcuts one way or the other.”

Nick Sullivan, at 23:40

Combining the two was the perfect fit.

Not a cryptanalyst

Jon asked a crazy question: did Nick ever sit in a coffee shop trying a new way to factor the product of two large primes? In the early days, before he knew the theory, Nick said, he played with it. Then he learned the quadratic sieve and the general number field sieve.

“it was like, oh, this is really deep.”

Nick Sullivan, at 24:36

The quadratic sieve is the second-fastest factoring method known, and the fastest for integers under about 100 digits. The general number field sieve is the most efficient classical algorithm known for integers larger than 10^100 (Wikipedia, Wikipedia). Cryptanalysis takes real creativity and deep knowledge, he said. At Apple he could protect content with encryption and work on authentication, device attestation, obfuscation and reverse engineering. At Cloudflare:

“I can upgrade the cryptography for 20% of the internet”

Nick Sullivan, at 25:46

He called the academics in cryptography brilliant. In his telling, there hasn’t been a new factoring algorithm in 20 years.

“I’m not the kind of sit in a room for seven years and solve the hardest problem in the world type of person.”

Nick Sullivan, at 26:08

He’s the kind who brings smart people together. Then, Jon added, you see what happens in practice.

Write it down

Once you figure something out, Nick said, explain it.

“One of the things that I emphasized in my job at Cloudflare is blogging.”

Nick Sullivan, at 26:30

He wrote dozens of technical deep dives. Once you learn something, it becomes second nature and harder to teach, so writing it down while it’s new reinforces your own understanding and brings people into the industry. He has mentored high school students who became interns and developers, and given guest lectures at universities. His LinkedIn lists Cloudflare blog posts such as a primer on elliptic curve cryptography and the launches of CFSSL and Red October (LinkedIn).

Society needs people in security who understand these things, he said, to make its technical infrastructure better than it is today. It works today, but “that’s not enough, I think.”

“That’s right. It’s not enough.”

Jon McLachlan, at 27:58

All of this is about people

Jon took it to leadership: you get to share what you know and pass it on, and at some point you give back to the community more than you take from it. Nick agreed, and Jon kept going.

“and at the end of the day, like, all of this is about people.”

Jon McLachlan, at 28:50

When a project loses sight of its impact on people’s lives, Jon said, he takes that as a red flag, steps back and asks whether it matters. Everyone sees issues with the world. People who can do something about them should.

“like to be the change that you want to see in the world”

Jon McLachlan, at 29:26

“Well, I know that, Jon, from your history.”

Nick Sullivan, at 30:01

Signal and MLS

There are a lot of ways to give back, Nick said. He had moved from full-time at Cloudflare to advising, and kept volunteering on the internet itself. He was co-chair of a couple of committees at the IETF, he said, including the Crypto Forum and Messaging Layer Security.

The Crypto Forum Research Group sits in the Internet Research Task Force, which runs long-term research groups and does not set standards (Wikipedia). Messaging Layer Security is a protocol for end-to-end secure group messaging.

Jon asked whether Signal is based on it. Signal came first, Nick said. It did end-to-end encryption between two people and scaled to groups with essentially a pairwise handshake, where every participant has to do the handshake with everyone. That doesn’t scale well. Signal brought forward security and post-compromise security: ways to recover when a key is compromised. MLS set out to keep those properties for groups of a hundred, a thousand, ten thousand, with key agreement that scales with the size of the group.

“And so we recently published an RFC on this”

Nick Sullivan, at 31:45

MLS was officially published on July 19, 2023, and it is designed for groups as large as 50,000 and for people on multiple devices. Its security properties include forward secrecy, post-compromise security and scalability (Wikipedia). Nick said Webex already used it. The Signal Protocol, developed by Open Whisper Systems in 2013, calls its ratchet self-healing: after an attacker compromises a session key, later messages are out of reach (Wikipedia). The calculator below counts what a pairwise design costs as a group grows.

Listen · Nick Sullivan on episode 38, from 31:20
About 25 seconds: Nick on scaling Signal's properties to big groups. The clip stops on its own. Read the full transcript.
Two cards side by side. Pairwise: every member does a handshake with every other member, and a group of 100 needs 4,950 sessions. MLS, filled: key agreement that scales with the size of the group, designed for groups as large as 50,000. Below, a list. Signal's properties to keep: forward secrecy and post-compromise security. Filled: MLS was published on July 19, 2023, and Google Messages and Apple Messages began rolling it out over RCS in May 2026. Then: "it doesn't scale very well." Nick, at 30:55
Pairwise handshakes against one group key agreement. Sources: the episode and Wikipedia.

One final document

Standards are another thing he’s passionate about, he said. Everyone building companies and products bumps into the same problems.

“And people have built the same solutions for the same problems in multiple different places.”

Nick Sullivan, at 32:21

Jon pleaded guilty, and Nick said he had done it too. To make something global and interoperable, he said, it helps to pool the work of a few experts into one final document.

He also reviews for top security conferences, even without a PhD, he said, after co-authoring over a dozen papers at Cloudflare, and the background reading keeps him fresh. He leans toward enabling the group: less final reviewer, more sanity check, keeping the process moving and the people involved.

“you can scale your impact by enabling others and sharing knowledge.”

Nick Sullivan, at 34:19

A secret group in a secret company

That came naturally over ten years in public, he said. At Apple, the team couldn’t be so public.

On his way out of Apple, Jon said, he looked for security roles inside the company, and found that security means different things to different teams. Apple takes corporate security seriously, and the two of them had worked in

“a secretive group inside of a secretive company that just was so secret.”

Jon McLachlan, at 35:04

The way Nick had learned to do technology is hard in a secret group within a secret org, he said,

“because to me, communication is so important.”

Nick Sullivan, at 35:20

Jon put it another way:

“security based on, like, keeping what you do a secret is not secure.”

Jon McLachlan, at 35:26

Secrecy raises the cost of an attack, he said, but it cuts out everyone who could help move the needle. It might solve a short-term business problem and still be misaligned with your values in the grand scheme of things.

Listen · Jon McLachlan on episode 38, from 35:22
About 35 seconds: Jon on why secrecy isn't security. The clip stops on its own. Read the full transcript.

The Heartbleed Challenge

Nick recalled the moment the power of the crowd became obvious to him: Heartbleed. He described a silly bug in OpenSSL, a widely used cryptography library, that let an attacker send a single packet and grab memory contents. Heartbleed was a missing bounds check in OpenSSL’s TLS heartbeat extension, introduced in 2012 and disclosed in April 2014. At disclosure, some 17%, around half a million, of the internet’s secure web servers certified by trusted authorities were believed vulnerable to theft of private keys, session cookies and passwords (Wikipedia). Cloudflare had patched ahead of time, he said, but the open question was whether someone could have taken the private keys for all those websites. The team spent about a day on experiments.

Someone out there had to have a better way, he said. So Cloudflare set up a vulnerable website and invited the world to take the key.

“And I remember the challenge, yes.”

Jon McLachlan, at 37:26

“we were kind of overconfident that it would take a while for people to do it.”

Nick Sullivan, at 37:38

In less than a day, he said, a dozen people had solved it, some in different ways. Wikipedia records that on April 12, 2014, at least two independent researchers stole private keys from the server Cloudflare set up. John Graham-Cumming’s write-up two weeks later says Cloudflare was “unsure if private keys could be accessed” when it launched, and that most solvers searched the leaked memory for the prime numbers at the heart of an RSA key (Cloudflare).

“So the power of the crowd is strong.”

Nick Sullivan, at 37:57

Listen · Nick Sullivan on episode 38, from 37:29
About 40 seconds: Nick on the day the crowd beat the in-house team. The clip stops on its own. Read the full transcript.
A timeline. 2012: the bug, a missing bounds check ships in OpenSSL's heartbeat extension. April 7, 2014: disclosure, a fixed OpenSSL comes out the day Heartbleed goes public. April 2014, filled: the challenge, Cloudflare, unsure whether keys could leak, puts up a vulnerable server and invites the world. April 12, 2014, filled: the key, at least two independent researchers steal its private key. Then: "So the power of the crowd is strong." Nick, at 37:57
The Heartbleed Challenge, from patch to stolen key. Sources: the episode, Wikipedia and Cloudflare.

Collaborate, mostly

Jon drew the lesson: we all face the same huge set of problems, and collaborating on the solutions helps everyone make the world better much more quickly. Nick added a caveat.

“It doesn’t always work that way, though, right?”

Nick Sullivan, at 38:22

Sometimes you have to make a strategic, bold announcement and take advantage of timing, he said. Jon agreed that some problems don’t have a one-size-fits-all approach.

Advising startups

Jon asked how advising was going. Nick was a technical advisor to a number of small companies. Since leaving full-time work at Cloudflare, he said, he had been exploring the up-and-coming startups, security among them. If a company solved a problem he saw in ten years at Cloudflare and with its customers, he was happy to work through it with the founders. He did technical advising and some matchmaking, for technical roles or investors.

“It’s obviously kind of a tricky time in terms of fundraising”

Nick Sullivan, at 40:12

Nobody knows what happens macro, he said, and the foundations still matter: product market fit, and something novel in security or other functionality.

He loves entrepreneurs and admires anyone who takes on the world to build something of their own. Having him on a pitch deck as an advisor helps founders raise money too, he said. Jon could picture the investors’ reaction:

“oh, it’s Nick? Nick from Cloudflare?”

Jon McLachlan, at 41:33

Trust is the currency

A network helps connect the dots and close deals, Jon said, and then made a bigger claim.

“People talk a lot about currency and the security space, but our main currency, I think, is trust.”

Jon McLachlan, at 41:48

Trust builds up over time, he said, and it’s “shared between people, not companies.”

Nick added that security is a diverse group of professionals and even non-professionals, whose communities don’t always overlap: corporate security, AppSec and applied cryptography among them.

Three prongs

Jon asked for predictions about the next huge change in cryptography, and offered his own: he thinks

“quantum is going to break everything.”

Jon McLachlan, at 42:35

Nick saw three different prongs, all exciting at the time, and Jon had touched on each. The first is post-quantum cryptography, being rolled out in a lot of places. Jon expected it in FIPS in about a year and a half. NIST finalized the first three post-quantum FIPS standards in August 2024 (Wikipedia). Nick had been part of the standardization work for a long time.

“because it’s important to get out ahead of it.”

Nick Sullivan, at 43:10

“The threat with quantum computers is that someone in the future will go back and take your encrypted data and decrypt it.”

Nick Sullivan, at 43:14

So you start encrypting with future-proofed algorithms now, he said, whether a quantum computer comes in 10 years, 20, or never. That attack has a name, harvest now, decrypt later: storing encrypted data today and waiting for a hypothetical Q-Day when it can be read (Wikipedia). His LinkedIn says his teams ran the first widespread post-quantum deployment on the internet, with Google (LinkedIn). In 2019, Cloudflare and Google ran two post-quantum key exchanges in TLS 1.3 for 53 days, with Chrome builds on one side and Cloudflare’s edge on the other (Cloudflare). Kevin Kane laid out the harvest attack on episode 96, and our post on harvest now, decrypt later covers what to move first.

Listen · Nick Sullivan on episode 38, from 43:03
About 25 seconds: Nick on why post-quantum work starts before the quantum computer. The clip stops on its own. Read the full transcript.
Three cards side by side. Post-quantum, filled: encrypt with future-proofed algorithms now, before someone can decrypt what they stored. Privacy-enhancing tech: cryptography for what tracking provided, such as multi-party analytics. Confidential computing: homomorphic encryption, PIR, PSI and searchable encryption, so you trust less. Below: NIST finalized its first three post-quantum FIPS standards in August 2024. Then: "because it's important to get out ahead of it." Nick, at 43:10
The three prongs Nick named, with what each one does. Sources: the episode and Wikipedia.

Tracking without the tracking

The second prong is privacy-enhancing technologies.

“I think there is a sea change happening in terms of how the internet is monetized.”

Nick Sullivan, at 43:33

“Advertising and personal privacy and tracking are at odds.”

Nick Sullivan, at 43:38

There are proposals, he said, to use cryptography for some of what tracking provided, in a privacy-preserving way. He pointed to a project called DAP, which uses a multi-party computation system to collect analytics privately. Secure multi-party computation lets parties jointly compute a function over their inputs while keeping those inputs private (Wikipedia).

“This is just at the very beginning.”

Nick Sullivan, at 44:09

Trust less

The third prong is confidential computing, and homomorphic encryption is

“the superpower, holy grail of that field.”

Nick Sullivan, at 44:22

The performance still isn’t great, he said, but other advanced cryptography gives you useful pieces: private information retrieval, private set intersection, searchable symmetric encryption. Homomorphic encryption runs computations on encrypted data without decrypting it first. Private information retrieval fetches an item from a database without revealing which one. Private set intersection lets two parties compare encrypted sets and learn only what they share (Wikipedia, Wikipedia, Wikipedia). Today, he said, it’s tied together with trust, contracts and promises.

“You don’t have to trust as much.”

Nick Sullivan, at 45:13

“this gives you something cryptographically verifiable”

Nick Sullivan, at 45:18

Jonathan Mortensen made the same case for AI on episode 75: a promise in a contract isn’t enough.

Listen · Nick Sullivan on episode 38, from 44:53
About 30 seconds: Nick on trading promises for something you can verify. The clip stops on its own. Read the full transcript.
Three boxes joined by arrows. Trust, a promise. Contracts, an agreement. Cryptography, filled, you can verify. Below, a list. Homomorphic encryption: computations on encrypted data, without decrypting it first. Private information retrieval: fetch an item from a database without revealing which one. Private set intersection: two parties compare encrypted sets and learn what they share. Filled: searchable symmetric encryption, also on Nick's list. Then: "You don't have to trust as much." Nick, at 45:13
From a promise in a contract to something you can verify. Sources: the episode and Wikipedia.

Phone or cloud

Jon had heard one approach: keep all the data on the device, since devices keep getting more powerful, memory cheaper and silicon cheaper. Nick pointed to AI. New machine learning and LLM applications are humongous, he said, and they don’t fit on phones anymore.

“So you’re going to have to have a hybrid architecture no matter what”

Nick Sullivan, at 46:02

Jon agreed: the models will be ginormous, the computers that run them will sit in the cloud, and there are lots of reasons to use a centralized SaaS service.

“And to be able to do that, like, and preserve your privacy at the same time, is ideal.”

Jon McLachlan, at 46:17

“it’s going to be a talking point.”

Nick Sullivan, at 46:30

On episode 75, Jonathan walked through Apple’s reference architecture for doing exactly that.

Three cards side by side. On the device: devices keep getting faster, and memory and silicon cheaper, so keep the data there. In the cloud: big models need machines built for them, and there are reasons to use a centralized SaaS service. Hybrid, filled: Nick, a hybrid architecture no matter what, if you do anything with AI. Then: "So you're going to have to have a hybrid architecture no matter what" Nick, at 46:02
On the device, in the cloud, or both, as Jon and Nick talked it through. Source: the episode.

Ideas from operators

Jon asked what doesn’t exist that someone should build. Nick, on the inspire-your-audience segment, didn’t have one idea to put on the table: there are so many. Many of the companies he advises started from problems he and his Cloudflare colleagues ran into.

“a lot of the best ideas come from people who are just banging their head against the walls inside of a big company”

Nick Sullivan, at 47:15

The companies that inspire him come from operators, he said: people with a generic problem they know isn’t being solved anywhere else, who go out and solve it.

Follow your curiosity

Jon asked what advice Nick would give his younger self. It’s a hard question, Nick said, and he’s relatively happy where he is.

“I feel like I’m much more of someone who learns through experience and through making mistakes”

Nick Sullivan, at 48:25

As a young person he was always chasing novel ideas and things that change how people interact with technology.

“be curious, my young self”

Nick Sullivan, at 49:16

Curiosity opens up the world, he said, and helps you know yourself.

Listen · Nick Sullivan on episode 38, from 48:52
About 40 seconds: Nick's advice to his younger self. The clip stops on its own. Read the full transcript.

What happened after the tape

Nick’s next roles. His LinkedIn shows his advisory contract with Cloudflare ending in January 2024, part-time work as head of research at Blind Insight, a searchable-encryption startup he co-founded, from October 2023 to March 2026, a visiting fellowship at the Center for Democracy & Technology in 2025, and a seat on the Internet Architecture Board from March 2026. He still co-chairs the Crypto Forum Research Group (LinkedIn).

Post-quantum went default. NIST finalized its first three post-quantum FIPS standards in August 2024. Hybrid key agreement that pairs classical ECDHE with ML-KEM, standardized as FIPS 203, has been on by default in Chrome since version 131 in November 2024 and in Firefox since version 132, and OpenSSL 3.5, BoringSSL and Cloudflare support it (Wikipedia).

Privacy Pass became a standard. The IETF published the Privacy Pass architecture, authentication scheme and issuance protocols as RFCs 9576, 9577 and 9578 in 2024 (Cloudflare Research).

Tampering got a dashboard. On September 5, 2024, Cloudflare launched a dashboard and API on Cloudflare Radar for connection tampering signals, built on the SIGCOMM work (Cloudflare).

MLS reached phones. In March 2025, the GSMA said version 3.0 of the RCS Universal Profile would support MLS, and Apple said it would support that standard. Google Messages and Apple Messages began rolling out MLS end-to-end encryption over RCS in May 2026 (Wikipedia).

The hostname went dark. Encrypted Client Hello, which encrypts the whole Client Hello including the name of the site, was defined in RFC 9849 in March 2026 (Wikipedia). Nick’s LinkedIn lists him as a co-author of ECH.

A startup he backed. His LinkedIn lists RunReveal among his exits as an investor and advisor. RunReveal announced it was joining ClickHouse on September 1, 2026, as covered on episode 50.

His post on a Berkeley talk. On October 29, 2026, Nick announced a talk at the UC Berkeley Security Seminar, set for November 16, on how a cryptographic construction becomes Internet infrastructure, with the CFRG’s work on post-quantum KEMs as the example. Earlier that month, he co-chaired an IAB workshop on deploying post-quantum authentication (LinkedIn).

Nick on how a cryptographic construction becomes Internet infrastructure.

Nick Sullivan · October 29, 2026
His post on the Berkeley talk, with links to the CFRG drafts. View on LinkedIn.

More on this show. Kevin Kane talked about cryptography and the quantum clock on episode 32. Riad Wahby of Cubist argued that security against speed is a false choice on episode 69. On the blog, Harvest Now, Decrypt Later covers which data to move first.

A timeline. 2024: Privacy Pass, RFCs 9576, 9577 and 9578. August 2024: FIPS 203, NIST finalizes its first three post-quantum standards. September 5, 2024: tampering on Radar, Cloudflare publishes connection tampering signals. November 2024: Chrome 131, hybrid post-quantum key agreement on by default. March 2026, filled: two milestones, ECH becomes RFC 9849, and Nick joins the Internet Architecture Board. May 2026, filled: MLS over RCS, Google Messages and Apple Messages begin the rollout. October 29, 2026, filled: a Berkeley talk, his post on how cryptography becomes infrastructure.
After the tape, from FIPS 203 to Nick's 2026 post. Sources: Wikipedia, Cloudflare, Cloudflare Research and LinkedIn.

Did you catch it? A seven-question quiz

Seven questions from the episode. Each answer links the moment in the transcript.

Quiz · seven questions

Did you catch it?

Pick an answer to see whether you caught it.

  1. What did Nick study at Waterloo?

    c. "I studied pure mathematics, actually." Transcript, 0:33.

  2. How did Nick describe what Cloudflare does?

    a. "it basically sits as a proxy between people and websites or web resources." Transcript, 3:03.

  3. What did Nick compare Cloudflare's free customer base to?

    b. "So it's kind of like an immune system, if you will." Transcript, 11:26.

  4. How long did most successful research projects at Cloudflare take to mature?

    c. "three to five years to become fully robust." Transcript, 13:21.

  5. What slogan did Nick come up with for the research team?

    a. "one of the slogans that I came up with for the research team was, we make friends." Transcript, 16:34.

  6. How did Cloudflare find out whether Heartbleed could leak private keys?

    b. "So what we did is we came up with this thing called the Cloudflare Heartbleed Challenge." Transcript, 37:21.

  7. What did Nick call homomorphic encryption?

    a. "the superpower, holy grail of that field." Transcript, 44:22.

Does your security make friends?

Eight questions on whether people bring your security team in, and whether your security holds up in the open. Each tip quotes the tape or paraphrases it, and links its section.

Scorecard · eight questions

Does your security make friends?

Answer for your company as it runs today. If two answers fit, pick the lower one.

  1. When engineers start a new feature, when does security hear about it?

    • After it ships
    • Right before launch
    • At the design stage, because they ask us

    Jon's second kind of security person is the one the room is glad to see. See Two kinds of security people.

  2. How do engineers reach someone with security expertise?

    • They file a ticket and wait
    • They know who to ask
    • A five-minute chat, any day

    A five-minute talk with an expert, Nick said, can send a team in a different direction. See A resource or a burden.

  3. When you need a cryptographic protocol, what do you use?

    • Something we built ourselves
    • A library, set up by whoever got there first
    • A published standard, such as TLS 1.3 or MLS

    Nick's case for concentrating the experts on one final document. See One final document.

  4. Does any part of your security depend on keeping the design secret?

    • Yes, it's the main defense
    • Some of it
    • No, it holds up if the design is public

    "security based on, like, keeping what you do a secret is not secure." See A secret group in a secret company.

  5. Can an outside researcher report a bug and get an answer?

    • There's no way to reach us
    • An inbox nobody watches
    • A published policy, and a reply within days

    "So the power of the crowd is strong." See The Heartbleed Challenge.

  6. Do you write down what you learn and share it?

    • It stays in people's heads
    • Internal notes, sometimes
    • Write-ups the whole company can read, some of them public

    "One of the things that I emphasized in my job at Cloudflare is blogging." See Write it down.

  7. Have you started moving data that must stay secret for years to post-quantum key exchange?

    • We haven't looked
    • We know where that data lives
    • Hybrid key exchange is on wherever our stack supports it

    Nick's reason to move now: someone can decrypt today's traffic later. See Three prongs.

  8. Could a vendor that holds your data read it?

    • Yes, we rely on the contract
    • Some of it is encrypted with keys we hold
    • Sensitive data is protected in a way we can verify

    "You don't have to trust as much." See Trust less.

Roadblock after roadblock

People route around security, and some of it depends on secrets. Start with a standard in place of homegrown crypto, and a way for outsiders to report bugs.

Healthy tensions

The pieces are there. Next, get into design reviews early, write down what you learn, and find the data that must stay secret for years.

We make friends

Engineers bring you in, and your security holds up in public. Keep the post-quantum work moving, and trade contract promises for things you can verify.

How many handshakes does a group need?

Nick described the first way to scale two-person encryption to a group: a pairwise handshake between every pair of members. This calculator counts what that costs.

Calculator

How many handshakes does a group need?

Two numbers: the people in the group, and the messages each one sends a day.

4,950pairwise sessions to set up
99encrypted copies of every message
198,000encryptions a day across the group

Pairwise sessions for a group of 100: 4,950. Encrypted copies of each message: 99. Encryptions a day, at 20 per person: 198,000.

Pairwise means every member keeps a session with every other member, so the count grows with the square of the group. MLS was built so key agreement scales with the size of the group, Nick said, and Wikipedia describes it as designed for groups as large as 50,000. The defaults are placeholders: put in your own group.

From pure math to the Internet Architecture Board

Nick’s path, the protocols he talked about, and the episodes on this site that pick up the thread. Filter by topic.

Timeline

From pure math to the Internet Architecture Board

  1. CISaC

    Cryptography researcher on elliptic and other algebraic curves, during his master's (LinkedIn).

  2. Symantec

    Security analyst, and a co-author of the Internet Security Threat Report (LinkedIn, on this page).

  3. Apple

    The DRM team, where his path crossed Jon's (LinkedIn, on this page).

  4. Cloudflare

    Head of security engineering, when the company had under fifty people (LinkedIn, on this page).

  5. Heartbleed

    The OpenSSL bug is disclosed, and Cloudflare's challenge server gives up its key within hours (Wikipedia, Cloudflare, on this page).

  6. Universal SSL

    Cloudflare makes SSL free for every site (Cloudflare, on this page).

  7. Head of cryptography

    Reporting to the CTO, on everything at Cloudflare that touches cryptography (LinkedIn).

  8. Privacy Pass

    The paper on anonymous tokens for skipping challenges, with Nick among the authors (Cloudflare Research, on this page).

  9. TLS 1.3

    RFC 8446 defines the version Nick worked on (Wikipedia, on this page).

  10. CFRG co-chair

    Co-chair of the Crypto Forum Research Group (LinkedIn, on this page).

  11. Head of research

    He founds Cloudflare Research (LinkedIn, on this page).

  12. The post-quantum experiment

    Cloudflare and Google test two post-quantum key exchanges in TLS 1.3 for 53 days (Cloudflare, on this page).

  13. iCloud Private Relay

    Apple announces the two-relay privacy service (Wikipedia, on this page).

  14. MLS

    Messaging Layer Security is published (Wikipedia, on this page).

  15. Advisor

    From head of research to an advisory contract with Cloudflare (LinkedIn).

  16. Episode 32

    Kevin Kane on cryptography and the quantum clock (episode page).

  17. SIGCOMM

    The connection tampering paper is presented in New York (Cloudflare Research, on this page).

  18. Episode 38

    Nick on the show (Podbean, transcript).

  19. Episode 50

    Alan Braithwaite of RunReveal, whom Nick interviewed at Cloudflare (episode page).

  20. FIPS 203

    NIST finalizes its first three post-quantum standards (Wikipedia, on this page).

  21. Tampering on Radar

    Cloudflare publishes connection tampering signals (Cloudflare, on this page).

  22. Chrome 131

    Hybrid post-quantum key agreement on by default (Wikipedia).

  23. Episode 61

    Taher Elgamal, the father of SSL (episode page).

  24. Episode 75

    Jonathan Mortensen on using AI without giving up your data (episode page).

  25. Internet Architecture Board

    A two-year term on the IAB (LinkedIn, on this page).

  26. ECH

    Encrypted Client Hello is defined in RFC 9849 (Wikipedia, on this page).

  27. MLS over RCS

    Google Messages and Apple Messages begin rolling out MLS encryption (Wikipedia).

  28. Episode 96

    Kevin Kane on the harvest attack (episode page).

  29. A Berkeley talk

    Nick on how a cryptographic construction becomes infrastructure (LinkedIn, on this page).

What to do this week

Five things, in the order of the episode.

  1. Take the scorecard. Ask whether engineers bring your security team in early, then take the scorecard.
  2. Use the standard. Swap homegrown protocols for TLS 1.3, and for MLS where you encrypt group messages, as in One final document.
  3. Give the crowd a way in. Publish how an outside researcher can report a bug, and answer within days, as in The Heartbleed Challenge.
  4. Start post-quantum with long-lived data. Find the data that must stay secret for years, and turn on hybrid key exchange where your stack supports it, as in Three prongs.
  5. Get help building it in. Our Product Security work puts security into the design. Our Anti-Reverse Engineering work protects the software you ship. If you want to talk it through, book 15 minutes with us.

Key takeaways

  1. Research runs on relationships

    Without short-term deadlines, a research team can't promise deliverables in six months. It needs people who know its value and want to work with it, hence the slogan: we make friends.

  2. Be the security person people want in the room

    One kind of security person brings roadblock after roadblock. The other has the same expectations and helps the team build the thing. A security person who isn't a resource for the company is potentially a burden.

  3. Give long bets five years

    Most successful research projects at Cloudflare took three to five years to mature. Nick counted about five from Privacy Pass to iCloud Private Relay, and the mission made the work fundable.

  4. The server side can see tampering

    A device in the path reads the site name in the handshake and sends a reset. Signatures in the packet headers let a big CDN measure tampering passively, around the world.

  5. A free tier works like an immune system

    Free customers bring threats into view early. Study them, build a mitigation, and protect the large enterprise customers.

  6. Standards spare everyone the same work

    Everyone bumps into the same problems and builds the same solutions. Concentrating experts on one final document makes things global and interoperable, as MLS did for group messaging.

  7. Secrecy is not security

    Keeping the design secret raises the cost of an attack and cuts out everyone who could help. Cloudflare's Heartbleed Challenge fell to outside researchers within hours.

  8. Move to post-quantum before you need it

    Someone can store encrypted data now and decrypt it once a quantum computer exists. Start using future-proofed algorithms now.

  9. Trade promises for proof

    Homomorphic encryption, private information retrieval, private set intersection and searchable encryption replace promises in a contract with something you can check cryptographically.

Chapters

Tap a chapter to play from there.

Pull quotes

“I studied pure mathematics, actually.”

Nick Sullivan ·

“Apple Media Empire is built on the foundations of having secure ways of enforcing business logic.”

Nick Sullivan ·

“I joined when it was a small company, under 50 people or so.”

Nick Sullivan ·

“participated in TLS 1.3, invented stuff like Privacy Pass.”

Nick Sullivan ·

“There's no authentication on any of these TCP things.”

Nick Sullivan ·

“So it's kind of like an immune system, if you will.”

Nick Sullivan ·

“This is a five-year process.”

Nick Sullivan ·

“Cloudflare's mission statement is to help build a better internet.”

Nick Sullivan ·

“And so, like, one of the slogans that I came up with for the research team was, we make friends.”

Nick Sullivan ·

“And Nick is the guy who's going to help us understand and navigate that complex space as an ally.”

Jon McLachlan ·

“It was kind of a series of events and doors that opened up.”

Nick Sullivan ·

“you can take data and make it so that no one in the world can see it.”

Nick Sullivan ·

“Everybody's taking shortcuts one way or the other.”

Nick Sullivan ·

“I'm not the kind of sit in a room for seven years and solve the hardest problem in the world type of person.”

Nick Sullivan ·

“you can scale your impact by enabling others and sharing knowledge.”

Nick Sullivan ·

“So the power of the crowd is strong.”

Nick Sullivan ·

“People talk a lot about currency and the security space, but our main currency, I think, is trust.”

Jon McLachlan ·

“The threat with quantum computers is that someone in the future will go back and take your encrypted data and decrypt it.”

Nick Sullivan ·

“Advertising and personal privacy and tracking are at odds.”

Nick Sullivan ·

“You don't have to trust as much.”

Nick Sullivan ·

“be curious, my young self”

Nick Sullivan ·

About the guest

Nick Sullivan, Owner, Cryptography Consulting LLC

Nick Sullivan joined Cloudflare in 2013, when it had fewer than fifty employees, and spent about a decade there as head of security engineering, head of cryptography and the founder of Cloudflare Research. Before that he spent six years on Apple's DRM team, where he worked with Jon McLachlan, and co-authored Symantec's Internet Security Threat Report. He co-chairs the IRTF's Crypto Forum Research Group and the IETF working group behind Messaging Layer Security, and joined the Internet Architecture Board in March 2026. He holds a BMath in pure mathematics from the University of Waterloo and an MSc specializing in cryptography from the University of Calgary.

About the host

Jon McLachlan is co-founder of YSecurity, the on-demand cybersecurity team for startups, and of Cyberbase.ai. The Security Podcast of Silicon Valley is hosted by Jon McLachlan and Sasha Sinkevich. Jon McLachlan on LinkedIn ↗

Context and sources

  • Episode 38 on Podbean, March 9, 2024, with show notes on Nick's path from pure mathematics at Waterloo through Symantec, Apple and a decade at Cloudflare. Source ↗
  • Nick Sullivan's LinkedIn profile: founded Cloudflare Research and led cryptography at Cloudflare for a decade, contributed to TLS 1.3 and co-authored ECH, co-chairs the Crypto Forum Research Group and the MLS working group, and serves on the Internet Architecture Board. Source ↗
  • Nick Sullivan's LinkedIn experience: CISaC from 2004, Symantec from 2006, Apple from 2007 to 2013, Cloudflare from 2013 as head of security engineering, head of cryptography and head of research, then advisor to January 2024, Blind Insight, and the Internet Architecture Board from March 2026. Source ↗
  • Nick Sullivan on LinkedIn, October 29, 2026: a talk at the UC Berkeley Security Seminar on how a cryptographic construction becomes Internet infrastructure. Source ↗
  • Cloudflare Research: Global, Passive Detection of Connection Tampering, ACM SIGCOMM 2023, by Ram Sundararaman, Louis-Henri Merino, Kevin Bock, Marwan Fayed, Dave Levin, Nick Sullivan and Luke Valenta. Source ↗
  • Cloudflare Research: Privacy Pass, Bypassing Internet Challenges Anonymously, Proceedings on Privacy Enhancing Technologies 2018, by Alex Davidson, Ian Goldberg, Nick Sullivan, George Tankersley and Filippo Valsorda. Source ↗
  • Cloudflare Research: RFC 9576, The Privacy Pass Architecture, IETF, 2024. Source ↗
  • Cloudflare, September 5, 2024: A global assessment of third-party connection tampering, with a new dashboard and API on Cloudflare Radar. Source ↗
  • Cloudflare, April 27, 2014: Searching for The Prime Suspect, How Heartbleed Leaked Private Keys, by John Graham-Cumming, on the Heartbleed Challenge. Source ↗
  • Cloudflare, October 30, 2019: The TLS Post-Quantum Experiment, run with Google in Chrome Canary and Dev builds for 53 days. Source ↗
  • Cloudflare press release, September 29, 2014: Universal SSL makes SSL free for every site, almost doubling the encrypted web. Source ↗
  • Wikipedia, Cloudflare: founded July 26, 2009, its services act primarily as a reverse proxy between visitors and a customer's hosting provider. Source ↗
  • Wikipedia, Heartbleed: a missing bounds check in OpenSSL's TLS heartbeat extension, disclosed in April 2014, and private keys stolen from Cloudflare's challenge server on April 12, 2014. Source ↗
  • Wikipedia, Messaging Layer Security: published July 19, 2023, designed for groups as large as 50,000, and rolling out over RCS in Google Messages and Apple Messages from May 2026. Source ↗
  • Wikipedia, iCloud Private Relay: two relays, an ingress proxy run by Apple and an egress proxy run by another provider, announced in June 2021. Source ↗
  • Wikipedia, Transport Layer Security: TLS 1.3 defined in RFC 8446 in August 2018, and hybrid post-quantum key agreement on by default in Chrome since version 131. Source ↗
  • Wikipedia, Server Name Indication: the hostname travels unencrypted in the handshake, and Encrypted Client Hello was defined in RFC 9849 in March 2026. Source ↗
  • Wikipedia, Post-quantum cryptography: NIST finalized the first three post-quantum FIPS standards in August 2024. Source ↗
  • Wikipedia, Harvest now, decrypt later: storing encrypted data until it can be decrypted, a hypothetical date called Q-Day. Source ↗
  • Wikipedia, OONI: the Open Observatory of Network Interference, monitoring internet censorship since 2012. Source ↗
  • Wikipedia, Alfred Menezes: co-author of the Handbook of Applied Cryptography and a professor in Waterloo's Department of Combinatorics and Optimization. Source ↗
  • Wikipedia, Signal Protocol: developed by Open Whisper Systems in 2013, with forward secrecy and post-compromise security. Source ↗
  • Wikipedia, Homomorphic encryption: computations on encrypted data without first decrypting it. Source ↗
  • Wikipedia, Secure multi-party computation: parties jointly compute a function over their inputs while keeping those inputs private. Source ↗

Transcript

Automatic transcription. Jon McLachlan hosts this episode with Nick Sullivan, who led cryptography and then research at Cloudflare. There is no diarization on the tape, so speaker labels were assigned from the content and checked against reference recordings of each voice from this episode and by pitch, and a few short lines may sit with the wrong speaker. Names and terms were normalized where the recognizer misheard them (Jon McLachlan, Nick Sullivan, Cloudflare, Waterloo, Alfred Menezes, Symantec, CFSSL, TLS 1.3, Privacy Pass, IETF, OONI, SIGCOMM, iCloud Private Relay, Akamai, Fastly, Messaging Layer Security, Webex, Heartbleed, OpenSSL, FIPS, DAP). Words in [brackets] are best guesses where the audio is unclear. Answers are transcribed as spoken, with most repeated words and false starts kept. Short backchannels such as "yeah" and "right" are left out where they overlap the other speaker.

Read the full transcript 8,131 words

Jon McLachlan

Hello, everyone, and welcome to another episode of the Security Podcast in Silicon Valley. I'm here today with an awesome guest and a good friend, Nick Sullivan.

Nick Sullivan

Hi, Jon.

Jon McLachlan

How's it going?

Nick Sullivan

It's going well. I'm really happy to be here. Thanks for having me on, and excited to chat with you.

Jon McLachlan

Just for our listeners, Nick has built an entire career around security and cryptography over the many years. You studied at Waterloo. Was that a focus on cryptography and security?

Nick Sullivan

I studied pure mathematics, actually. So it has a connection to security and cryptography, but it is much more abstract and much more theory focused.

Jon McLachlan

Well, you got to get those foundations solid before you can get into the bit smashing, right?

Nick Sullivan

True.

Jon McLachlan

So when you entered the industry, it looks like you were a cryptography researcher for a while, and then you jumped into Symantec as a security analyst, and then you were at Apple as a [mathemagician]. I love the title. We worked together, just for all of our listeners' context. We were both on the DRM team. Nick worked a little bit more on the cryptography, the actual business logic, and I was over in the harder to talk about obfuscating compiler team, but we worked very closely together. Those were the glory days, weren't they, Nick?

Nick Sullivan

Yeah, yeah, that was fun. Apple Media Empire is built on the foundations of having secure ways of enforcing business logic. And that's what DRM is.

Jon McLachlan

I know, that's what DRM is, yeah. There's a very strong, like, political opinion out there on if it's good or if it's evil. Maybe it's both. But we both worked on it. So without even really having to dive into the politics behind it, the technical challenges were very open and interesting. But Nick had the good sense to get out of there and join a young, budding, promising Cloudflare, which you have spent more than 10 years there now. And you started off as head of security engineering, and very quickly that turned into head of cryptography. You facilitated a lot of really great public talks, which I actually participated in, which was amazing. And you transformed into the head of research there at Cloudflare. And you are still associated with Cloudflare as one of their advisors, which is spectacular. And now you have a cryptography consulting LLC, along with many advisor roles to lots of promising companies.

Nick Sullivan

Yeah, yeah, it's, you know, Cloudflare was quite an amazing adventure. I joined when it was a small company, under 50 people or so. And what Cloudflare does is it basically sits as a proxy between people and websites or web resources. And just that general, generic framework makes it incredibly powerful. So you can do things like security monitoring, you can use web application firewalls, you can stop denial of service, you can do all sorts of things if you're the middle party between the user and the website. And as we know, the web is constantly, constantly under attack. Constantly, attackers are finding new ways to exploit technical loopholes to try to take advantage of companies, take advantage of people, stealing identities, stealing money. You know, the world is a wild and crazy place. And having some sort of security blanket over your website and web services is really helpful. And so, you know, when I started there, I was doing some engineering, some pretty hardcore engineering, solving some of the bigger, you know, security architecture things that were missing. So I built Cloudflare's public key infrastructure and built a small team to open source tool called CFSSL for creating certificates and managing all sorts of internal trust things. And yeah, it really helped what was a fledgling company at the time take off. And so it's been an amazing journey there for 10 years. And, you know, I can go a little bit more into some of the different things that I touched there, but it was an amazing opportunity to be able to take the ideas of security and cryptography and privacy and all these wonderful things that the tools, that the technical tools that we build and work on create, and bring it to a wide audience. And not only bringing it to a wide audience as part of a product, but I spent a lot of time working on the industry challenges themselves directly. So I worked with standards bodies, I've worked with academia on validating new ideas, built prototypes, I worked with all the folks at Cloudflare to kind of bring these exciting new ideas that were nascent in the industry and explored in academia, but not to, from like its academic rigor, but not necessarily from a technical scaling perspective, able to be deployed, and kind of help bridge that gap. And that started with my own personal history and focus is on cryptography things, and that's where the head of cryptography role is. We participated in TLS 1.3, invented stuff like Privacy Pass. And, you know, that model of taking less of a product-focused approach and more of a research, long-term technology approach that is based on collaboration across the industry and not competition, through venues like the IETF, through conferences, through working with academics. And we built up this amazing internship program, with probably 40 interns at least going through the program over the last couple of years. And this made a lot of sense to go beyond security and cryptography. And so we expanded the team from research into, like, cryptography and protocols to broader research. So we brought in things like DNS, network measurement, distributed systems, privacy-enhancing technologies. And yeah, I still consult with Cloudflare, and the team is still going strong, putting out some amazing stuff. We recently published a paper from my work there, studying basically the effect of, it could be called censorship, but it's more broadly, if you were to categorize, as tampering, as different participants on the network stopping different traffic from getting through in various parts of the world. And we had some pretty interesting insights there, where, you know, almost 10% of the traffic that's, or the requests that were intended for Cloudflare's millions and millions of customers, that makes up almost 20% of the web at this point, 10% of those at least were getting [blocked] for one reason or the other, in really global. And it's a lot higher in certain countries. And so there's a lot of different reasons you could imagine that someone would want to stop a connection. There's corporate security, there's lots of monitoring protection there so that employees don't go to phishing sites and this sort of thing. But there's also geopolitical reasons, and there's things like schools, like schools that have content filtering mechanisms that come into play. So being able to be at Cloudflare, which is central to a big percentage of the web, gave us some interesting, amazing data.

Jon McLachlan

Right? Yeah, you have just tons of data, I imagine. So were there any patterns that really stuck out, that struck you as unexpected?

Nick Sullivan

Patterns with respect to, like, product development, or with, like, what has worked and what didn't work?

Jon McLachlan

Oh, I mean, the censorship.

Nick Sullivan

Oh, [and] the censorship?

Jon McLachlan

How could you detect that there was something being censored, like, within a school? You know, that's tricky to be able to detect, right? Because if the packets are not making it to Cloudflare's CDN or connection points, like...

Nick Sullivan

Yeah, no, it's a good question. It's very counterintuitive that one would be able to learn this from only having the server side of the connection. And historically, up until this point, a lot of the measurements with respect to censorship and with respect to connectivity and availability of content online has been active, meaning that companies and groups like OONI and these other measurement groups who try to measure these things are putting devices in people's homes around the world, or, like, letting people opt in. And then the measurements would come to, like, a preset list of names, just to see whether or not these ones were available or not. And with our study, the folks at the University of Maryland came up with this brilliant machine learning algorithm to analyze TCP flows. And what happens is that there's a number of different signatures that you can see on the server side, where one packet will get through and it'll be indicative of the entire flow being blocked or delayed or, you know, whatever. Like, from a very high level, for those who understand TCP, if you get an inbound SYN, then you can just send a FIN if you're in between, right? There's no authentication on any of these TCP things. So someone will send a request, it'll have a website name in what's called the subject name indication. It's part of the raw records on the wire. It's like, I want to go to example.com, and it's in the handshake. You can see it, having, like, a network monitoring device. And if you match it against, you know, no-no words, you can send a reset packet, or you can do all sorts of different things just to confuse the state machines on either side. And so, because, like, the first request, maybe not the, I mean, maybe the handshake will get through, or... And that's how we can determine which sites are being tampered with, is that the website handshake goes through, it says, I'm going to this site, and then you get a random, you know, reset or a different packet that just tells the TCP connection to shut down. And so that's a strong indicator that something's going on. And there's a lot of signatures that are pretty high confidence, and correlated with particular products and devices that people sell that do this sort of thing. And so, yeah, the paper was just presented at SIGCOMM, a huge ACM conference. And yeah, that's just one of a million different things you can do with this amazing global data set. And I always say that's one of the superpowers of Cloudflare, is having a free customer base, having the ability for millions and millions of customers to just join on and use the service. And they get a lot of value out of that. And there is cost to Cloudflare, but Cloudflare gets a lot of value in terms of being able to see what threats are going on on the internet. So it's kind of like an immune system, if you will. And you see some sort of threat against a free customer, and then you can study it, apply mitigation, and then protect the very large enterprise customers. It's a really good symbiotic relationship.

Jon McLachlan

So there's value being added on both sides, unlike Google, where they're using the data to sell something.

Nick Sullivan

Yeah. And some of the work that I did most recently at Cloudflare was only facilitated by the fact that the company is not based on monetizing individuals, or trying to track people and sell information, all these sort of things. The business is based on, let's just make the network work the way that it should. And so in doing that, it enabled us to do a bunch of projects. Actually, one was with our former employer, Apple. Cloudflare participated in iCloud Private Relay, which is a privacy proxy that leverages CDNs and network providers like Cloudflare and Akamai and Fastly to provide a layer of anonymity for the client IP address. And it's native on iOS now. And that kind of happened through five years from when we came out with Privacy Pass and published it. A lot of behind-the-scenes conversations, convincing people of the benefit of secure DNS and privacy and all these sort of things, where Apple and Cloudflare were very aligned. Neither company is trying to monetize individuals, so they can build these more robust privacy primitives and privacy protocols to help improve user privacy. And Cloudflare was happy to do that. And that's what I would consider the kind of wins that you have to expect when you have a research team, and also the timelines that you have to be comfortable with as an executive team funding something like a research team. This is a five-year process. And actually, most of the successful projects at Cloudflare took, you know, three to five years to become fully robust.

Jon McLachlan

Did you find that that was a difficult sell, when you had to interact with the executive teams and pitch these projects and the potential payout from a business sense for Cloudflare, but also in terms of, like, let's just make the internet better and more secure and more private?

Nick Sullivan

Yeah, the mission statement was very much in support of doing these sort of things. Cloudflare's mission statement is to help build a better internet. And so better means a lot of different things. It can mean secure, more available. It can also mean more private. And so, you know, Cloudflare was fantastic in supporting myself and the research team in doing these long-term things in pursuit of the mission, rather than in pursuit directly of the product. And, you know, research teams are a cost center, and you have to be aware of that when building a company. Whenever you're growing as fast as Cloudflare has for the last 10 years, consistently growing significantly year over year, there's always a million types of tensions that have to come into play.

Jon McLachlan

Healthy tensions.

Nick Sullivan

Yeah. That's how things are funded, how things are prioritized, and how people communicate, especially as you go from 100 people to 500 people to 2,000 people. It's very different systems that have to be in place. But given the strong support from leadership that, you know, Cloudflare was serious about the mission of helping build a better internet, these long-term projects, although, like, lean as they were, were very successful. So that was really fun. And I think it's, it's something that not a lot of companies are comfortable taking a big swing at, is long-term research, because just how precarious the situation is with raising revenue and becoming cash flow positive and all these sort of things that businesses have to deal with, making sure that growth continues at a rate that's, if it's not exponential, that it's, it's, it's repeatable and in line with the investor's goals. But I think we showed that research can pay off, even in a smaller company, even in a growing company, and took the model, took some lessons from different models at research groups at places like Google or Microsoft, or even going way back to the Bell Labs days. And yeah, it's a different way of operating than you will at a product-focused company, where it's very, you know, you have very short deadlines and sprints, and everything has to come together in terms of schedules. And having an aspect of your company that doesn't have clear short-term deadlines, meaning, like, within even six months, there may not be guaranteed deliverables or outcomes, it relies, well, it forces you to rely more on relationships, relationship building within the company, and communication, making sure that people know the value that research brings and will bring, and want to work with you. And so, like, one of the slogans that I came up with for the research team was, we make friends. And there's absolutely no way to do some of the more difficult, long-term things that take away from other priorities that come with research without human support and without people.

Jon McLachlan

Of course, of course. I feel the warmth, and, like, the Canadian, like, welcoming spirit when you said that to me. It's just like, and being in Minnesota, it's well received. Yeah, I like that. We make friends. I've always mentioned that there's two types of security people in the world. There's one type of security person, you walk into a meeting room and everyone looks at you and they go, oh, no, like, so-and-so is here, they're going to tell us we have to implement all these strange standards, and they're going to push the deadline back by, you know, a couple of months, and, like, oh, it's just going to be roadblock after roadblock. And then there's another type of security person. They walk into exactly that same meeting room, they have exactly the same expectations of what they want to see in the product, but the other people in that meeting, they see this person walk in the room and they're like, oh, I'm so glad, like, Nick is here. He's going to help us understand how to build this new thing in a way that aligns with our security goals and our customers' expectations, and just make the world a better place for everyone, and maybe more private, more secure. And Nick is the guy who's going to help us understand and navigate that complex space as an ally.

Nick Sullivan

I'd like to think that I try to embody that second person, like...

Jon McLachlan

Perfectly and perfectly. And, like, even your slogan of, like, we make friends. Yes, I feel it.

Nick Sullivan

Yeah. Well, thank you. Yeah, I really don't think that it's possible to be very successful in a company if you're, if you're, if people's expectations coming into the meeting is, like, I don't want to see that person, they're going to cause trouble for me.

Jon McLachlan

So you've lost already, like, at that point.

Nick Sullivan

Yeah, of course. And so, you know, my approach is very collaborative, very inclusive, very helpful. Like, I encourage everyone on my team to, you know, as much as possible, if people are looking for help or you have some expertise that you can provide, help people out. I mean, it's going to help the company, and it's going to help the position of the goals of the group. And, you know, there's specialized knowledge within the research groups or security groups, where some people have years of experience and understand things in the ways that, you know, a typical engineer will just not know at all, and will make very easy-to-find mistakes, or go in directions where just, you know, a five-minute conversation with somebody who's an expert may help turn them in different directions. And if you're not a resource for your company, then you're potentially a burden, right? Especially if the goals of your group and the people that you're working with are perceived as being, like, against the goals of the company, right?

Jon McLachlan

Right. Yeah, that's right. It's just a dangerous place to be. Especially in a contracting market, I would say. But maybe that has nothing to do with how a company might contract its resources as well. Like, I'm sure both types of security folks are considered when you just have to contract the size of a company, with layoffs or whatnot. Seems to be going on a little bit more often these days than...

Nick Sullivan

Yeah, yeah, I mean, efficiency is super important. If you can do more with less, then everybody's going to want that.

Jon McLachlan

Yeah, for sure. For sure. So is there any story about how you got into security? I know you went down the mathematics road, and cryptography maybe piqued your interest, and it was a little bit more applied mathematics when you got into the cryptography space. But was there a moment, or, like, something hit your passion button that tipped you into that security space?

Nick Sullivan

You know, I wish I could identify a specific moment, but I don't think there was one. It was kind of a series of events and doors that opened up. You know, in my last semester at Waterloo, in my pure math degree, I'd done pure math, combinatorics, all this very pen-and-paper, deep, abstract stuff. And then Alfred Menezes, who was one of the cryptography profs at Waterloo, he actually invited me back to give a guest lecture a couple of years ago, which was super nice, taught this intro cryptography class, and everything just clicked for me, in that, like, I was able to, you know, take my deep, hard-learned intuition around prime numbers and discrete mathematics and all these sort of things, and put it into a concrete application. And cryptography to me is, it's like magic. Like, it allows you to provide things digitally that could only be reproduced physically. So, and then the intuition is really weird. So it's like, you can take data and make it so that no one in the world can see it.

Jon McLachlan

Right.

Nick Sullivan

And it's like, okay, well, we can physically do that by putting it in, like, an unbreakable safe, but you can do that digitally. And some of the, even the very basic things like digital signatures, they enable these capabilities that really help with designing the architecture of communication systems. And so after that, I decided to do a master's degree in cryptography, and expanded from math to doing computer science, did a couple courses, and got pretty involved with antivirus and exploits and security, like just the entry level into, like, okay, you know, there's a, there's a way that you think software works, and then there's a way that software really works. And these two things really tied together really nicely. And so once I finished my master's degree in elliptic curve cryptography, I, you know, started getting into some implementation, and then also studying things like threats and the internet. And Symantec had this great role for security analysts. They have an annual report called the Internet Security Threat Report, and yeah, they hired me as one of the co-authors of this. And this kind of opened the door to, like, massive data problems, and data as a tool for explaining the world of security. And so, you know, you have this cryptography angle, which solves security problems by unlocking these, like, magical primitives. And then there's just the fact that every single person, not every single person, but, like, the majority of the world is online and living their lives online and building things online. And there are security threats to this, and nothing was built perfectly. Everybody's taking shortcuts one way or the other. We're all just trying to make it all happen. And so that was just such a fascinating field for me, and combining those two was the perfect fit.

Jon McLachlan

Yeah. Perfect fit for Nick. That's amazing. Yeah, that's really cool that you came in from the mathematical side. Did you ever, did you ever, here's a crazy question for you, did you ever just, like, sit down at a coffee shop and be like, well, let's see if I can figure out a different way to think about how to factor the product of two large prime numbers, or [perform] a discrete [logarithm], like, efficiently over a modulo prime, or something like that?

Nick Sullivan

I mean, I definitely, in the early days, before I knew the theory, played around with things like this, right? What sort of algorithms could you do? But, you know, once I learned about the quadratic sieve and the general number field sieve, which are the most efficient ways of solving factorization over a certain size, it was like, oh, this is really deep. Like, this took all of my undergrad pure mathematics and algebra classes to, like, really understand what's happening. And so I think my creative juices, in terms of being a cryptanalyst, it was a little too, it involves, like, some real creativity and deep, deep knowledge of this, that, you know, I wasn't in the position to spend my life just [digging].

Jon McLachlan

Let go of that, and, like, go over here and do this industry stuff for a little bit.

Nick Sullivan

Yeah. Well, once it was clear that I could do something like at Apple, like, okay, let's protect all the content in the world with encryption, and do interesting things for authentication and for device attestation and for obfuscation and reverse engineering. Like, it's just such a tangible, physical landscape to me, at least. Like, when I think of software development, I mean, I think of it as, like, a physical thing. So it becomes something that you can explore, like a landscape. And so when exploring these sort of things at Apple, and then at Cloudflare, it's like, okay, I can upgrade the cryptography for 20% of the internet, and it's this amazing impact. And, you know, the academics in cryptography are amazing. They are so brilliant. And they spend their whole careers on things like, there hasn't been a new factoring algorithm for 20 years, right? And so, I don't know. I'm not the kind of sit in a room for seven years and solve the hardest problem in the world type of person. I'm very much a, let's work with people, to get, bring all of these amazingly smart people together.

Jon McLachlan

And, like, let's see what happens when the rubber hits the road.

Nick Sullivan

And then, once you do figure things out or there's new concepts, explaining it. You know, I'm a really big fan of communication. One of the things that I emphasized in my job at Cloudflare is blogging. So I wrote, you know, dozens of technical blog posts, just deep dives into these ideas, so that, you know, something that engineers tend to do, which is, I guess this is true, and this is true for everyone, and for any discovery and realization of new knowledge, is that once you learn it, it becomes more second nature, so it becomes harder to teach people how to learn what you learn. And so taking the time to break down something that you recently learned, that's new and innovative and novel, in a way that's either a general audience or a technical audience can follow, it not only reinforces your understanding of where this concept can be applied in other places, but it brings people in and brings people into the industry. And that's something I'm very passionate about. I've mentored high school students who then became interns and developers, and I have spoken at universities, I've given guest lectures at classes, and all these sort of things. I really think that this stuff is really fascinating to me, and a lot of people will find it fascinating, but it's also critically important for our society going forward to have folks working in security who understand these things and, you know, can help build up this society and this technical infrastructure to be better than it is today. Because it works today, but, you know, that's not enough, I think.

Jon McLachlan

That's right. It's not enough. I think there's a lot of interesting problems. And, like, as you mentioned, like, one of the great responsibilities, or, that's the right word, privileges of being a leader is that you get to share that, you get to pass it on, you get to communicate that. And it's part of being a leader. You get to a point in your career, you're giving back to the community much more than you're consuming from it. You're becoming a leader. You're right there at the tip of the [spear]. And to be able to mentor and then watch other people grow and realize, like, all of their potential, and then continue learning and growing, is a great privilege of being in a leadership position.

Nick Sullivan

Absolutely.

Jon McLachlan

It brings a different type of contribution back to the community, you know, not just technical, right? It's, and at the end of the day, like, all of this is about people. If we're not focused on how we can have an impact on people's lives and improve people's lives, it's like, I always try to take that as a red flag and, like, step back and ask, okay, how is this actually going to impact people, and does this matter, right? And if it does, then you can connect it back to people, in building a better world. Because we all see, like, there's issues with the world. Everyone sees issues with the world. And yeah, it's important to call those out. But I think it's even more important, especially for people that can do something about it, to do something about it, like to be the change that you want to see in the world, to help nudge the world in that direction. And depending on where you are in life and your career and the community, the communities that you're part of, like, you can take on different roles to do that. It sounds like you've just sort of implicitly picked that up and really value that as well. So that's special. And thank you. Thank you. Not everyone picks that up quite as quickly, or integrates that as a core importance, but I've always tried to as well.

Nick Sullivan

Yeah. Well, I know that, Jon, from your history. You're definitely someone who does what they can to help folks and to, you know, see the bigger picture. So I really appreciate that about you. And, you know, there's a lot of ways of giving back. I'm still, you know, although I've moved on from full-time at Cloudflare to being an advisor, I'm still involved in a lot of volunteer efforts to help improve the internet. So I'm the co-chair of a couple committees at the IETF, including the Crypto Forum and Messaging Layer Security, which is a new protocol for end-to-end secure group messaging.

Jon McLachlan

Is that what Signal is based on?

Nick Sullivan

Yeah, so Signal was sort of the OG, right? It was the original, we are able to do end-to-end secure encryption between two people. And they were able to scale it up to larger groups through, I guess, essentially a pairwise handshake. So it doesn't scale very well. Every participant of the communication has to, you know, chat with everyone.

Jon McLachlan

Do this handshake thing.

Nick Sullivan

Yeah, do this handshake thing. But it had some really nice, Signal introduced some really nice characteristics, like forward security, post-compromise security. So, like, if a key is compromised in the past or in the present, or it's currently compromised, there's, like, ways to recover, which is, which has been fantastic. And so MLS, Messaging Layer Security's goal was to build on these really strong primitives and security characteristics that something like Signal will bring in the two-party case, but to bring it to larger parties, a hundred, thousand, ten thousand, in a way that the key agreement can scale with the size of the group. And so we recently published an RFC on this, and it's used by Webex already for end-to-end secure communication. And that's something that I'm passionate about as well, is standards, in a way, not... finding a way to have a very polished and fully thought-out version of a tool to solve a particular problem, because we all run into similar problems in our building companies, building products.

Jon McLachlan

We do. Like, we all bump into the same problems over and over again.

Nick Sullivan

And people have built the same solutions for the same problems in multiple different places.

Jon McLachlan

Guilty as charged.

Nick Sullivan

Yeah. I mean, I've done that as well, in certain times in my history. But if you want things to be actually global and interoperable, it's great to concentrate the efforts of some experts to come up with a final document. And so I'm really big on that type of community, collaborative process for coming up with, like, standards and final documents there. I also spend some time helping with security conferences. So through my time at Cloudflare, I managed to work with a number of amazing people, co-authored over a dozen papers about interesting topics. And even though I don't have a PhD, I still was invited to help review some of these high-end conferences in security. And that keeps me fresh. Like, I really have to do the background reading to understand some of the latest papers that have been coming out. But I do enjoy taking what I've learned and my expertise and sharing it in ways that help improve the product of papers and standards. And I don't like to be pushy and take my design philosophy or particular way of doing things. I think every person has their own quirks, and we've all got our things. Yeah. So I've been more leaning into enabling the group, or people to participate in a group, and me being less of a final reviewer, but more of, like, sanity check, and making sure that the process goes on and that people are involved, and all these sort of things. So I think all that's super important. And, you know, as much as technical work is fun and it's great to just do it, like, you can scale your impact by enabling others and sharing knowledge. And I think that's just something that's come very natural to me over, well, I guess over my last 10 years of being in the public, whereas, you know, at Apple, we weren't allowed to really be that public.

Jon McLachlan

So, and, you know, like, when I was on my way out of Apple, I looked internally to Apple to see, like, what sort of security opportunities there were. And I talked to that global security team, and I talk with, like, ex-CIA, FBI, NSA people that were like, oh, yeah, here's what we do. And I was like, oh.

Nick Sullivan

There's, I mean, there's different, different types of security, right?

Jon McLachlan

There are different types of security. Yes. They take corporate security very seriously over there at Apple. It's a very, well, we worked in a secretive group inside of a secretive company that just was so secret.

Nick Sullivan

Yeah, it's really hard to do technology the way that I've learned has been effective at a place where you're in a secret group within a secret org, because to me, communication is so important.

Jon McLachlan

At the end of the day, I think another way to say it is that security based on, like, keeping what you do a secret is not secure. You increase the cost of the attack, but you're cutting out other people from being able to contribute and move the needle from the bigger picture. And you're not participating in, like, a global community. And sure, it might be a short-term, temporary solution to a very specific business problem or challenge that you face, but in the grand scheme of things, it could be totally misaligned with your values. And that's what I feel. We both shared that, like, in our time at Apple.

Nick Sullivan

I recall a moment in which the power of the crowd became very, very obvious to me, versus the power of, like, a red team or, like, an internal group reviewing things. And this is when the Heartbleed vulnerability came out, which was, I think, nine years ago now, and maybe 10 years ago. In any case, this was a bug in very popular cryptography library, OpenSSL, which was in almost every web server, like 80% of web servers in the world. And it had this really silly bug that was written by a grad student that allowed an attacker to send a single packet and then grab memory contents. Wild. This is, like, absolutely wild type of attack, and especially since it was enabled on all web servers. And so, you know, when I was at Cloudflare, we were internally debating, like, you know, we patched it ahead of time, but we were internally debating, like, if someone knew about this earlier, could they have gotten the private keys for all of these websites and all these customers? And it wasn't really clear. We tried to attack it ourselves. We spent, you know, like one day doing some experiments. And in the end, it was like, this is a lot of work. There's a, you know, this is a top news story in the world. Someone's got to have a better way to do this. So what we did is we came up with this thing called the Cloudflare Heartbleed Challenge.

Jon McLachlan

And I remember the challenge, yes.

Nick Sullivan

Yeah. We set up a vulnerable website and said, okay, world, you know, get the key from here. Let's see if you can do it. And a lot of really funky stuff happened, but we were kind of overconfident that it would take a while for people to do it. And in less than a day, there was, a dozen people had solved it, and some in very different ways. And so, I mean, at that point, it was like, okay, well, this is, there's no way we can get this much brain power from an in-house group, especially as a small company with limited resources. So the power of the crowd is strong. There are brilliant people out there, and if you can get them to pay attention to your problems, they might be able to help you.

Jon McLachlan

Absolutely. When you step back, there are a huge set of problems that we all face, and if we collaborate instead of compete over the solutions of those particular problems, like, we can all help each other make the world a better place much more quickly. Much more quickly.

Nick Sullivan

It doesn't always work that way, though, right? I mean, sometimes you have to make a strategic and bold announcement to take advantage of timing, and...

Jon McLachlan

That's true. Yeah, so there's a certain type of problem where it's not the one-size-fits-all sort of approach, of course.

Nick Sullivan

Right.

Jon McLachlan

And I know that you also give back, well, you also participate in the security community as an advisor for many startups too. So how's that going? What sort of advisor are you?

Nick Sullivan

Yeah, so I'm a technical advisor for a number of different small companies. And, you know, this has been great for me. So since moving away from Cloudflare, first full-time, they're doing great, I've been trying to explore more of what's happening in the up-and-coming startups. In security is one area, but basically everything that my expertise, or everything that I touched at Cloudflare, if there's a company doing something innovative there that solves a problem that I know, that I've seen through 10 years of working at Cloudflare and working with Cloudflare's customers, then I'm happy to talk with the founders and work through different problems. And so what I've been doing is some technical advising, as well as sort of matchmaking, if you will, for technical roles or investors or these sort of things. But yeah, I mean, there's security protocols and everything, sort of, in all sorts of these little companies that are really trying to solve important problems. And yeah, if I can be helpful and I love the idea, then, you know, this is sort of what I'm doing with my time right now, is helping these entrepreneurs succeed in whatever way I can. And so this has been really fun. I think there's a lot of opportunity in the industry. It's obviously kind of a tricky time in terms of fundraising and in terms of an economy. And yeah, yeah, like, to be more efficient. Like, what's going to happen macro, we don't really know. There, I mean, we could be in for a tough couple years. There might be, who knows. But in any case, having the right foundations for a company, finding product market fit, actually, you know, providing something novel from a security or other functionality. Yeah, it's been really great. And, you know, I love entrepreneurs. I really think that taking on the world and trying to build something by yourself is really laudable. And so, and I think they're often pretty fun people as well. So yeah, I've had a great time advising companies. And, you know, I'm actively talking to companies and talking to VCs about, you know, whether my advisorship, technical or otherwise, could be useful. And yeah, met some really great folks.

Jon McLachlan

Amazing. There we have it, for all the founders that are listening. Might have an interesting conversation with Nick. Should reach out.

Nick Sullivan

Yeah. It's also nice for them to have me on, like, the pitch deck for raising money, as an advisor and whatnot.

Jon McLachlan

I bet that helps significantly. Like, those VCs will just be like, oh, it's Nick? Nick from Cloudflare? Okay. All right. Good. Yeah. I imagine that your network is quite extensive as well. To, like, it just help connect the dots. And a lot of times that's what it takes to close a sales deal, or... People talk a lot about currency and the security space, but our main currency, I think, is trust. And that's built up over time, and it's shared between people, not companies. We evolved with those. To ignore them or to try to escape them is to almost deny our humanness.

Nick Sullivan

Yeah. It's a diverse, diverse group of professionals and even non-professionals working in security. And the communities are not always overlapping, too. So having exposure to folks who, well, corporate security versus AppSec versus cryptography in application, or...

Jon McLachlan

Well, what do you think will be the next huge change that come down the pipe in cryptography? Any predictions? I really think, I think quantum is going to break everything. Maybe homomorphic encryption will come pull through. We can encrypt, or operate on encrypted data, or something different, maybe.

Nick Sullivan

You know, I think there's three different prongs, and you touched on them all, that are exciting right now. One is post-quantum cryptography, which is being rolled out currently in a lot of places.

Jon McLachlan

My understanding is, like, it's going to be part of FIPS in about a year and a half.

Nick Sullivan

Yeah. So this is amazing. I've been part of this journey for standardizing post-quantum cryptography for a long time, because it's important to get out ahead of it. The threat with quantum computers is that someone in the future will go back and take your encrypted data and decrypt it. So, like, we need to start encrypting things with future-proofed algorithms now, in case, you know, 10 years, 20 years, who knows when a quantum computer, if and when it'll ever happen. But, you know, and the second one is privacy-enhancing technologies. I think there is a sea change happening in terms of how the internet is monetized. Advertising and personal privacy and tracking are at odds. And there are currently a lot of interesting proposals for using cryptography to enable some of the functionality that would have been provided by tracking, but in a privacy-preserving way. And this is happening in a lot of different venues. It's, yeah, there's a project called DAP, which uses a multi-party computation system to help collect analytics in a privacy-preserving way. This is just at the very beginning. I think it's going to be a huge wave going forward. And then the last is confidential computing, which you mentioned. Homomorphic encryption is, you know, the superpower, holy grail of that field. Still, the performance isn't that great, but there's other categories of types of things that do involve advanced cryptography that give you different useful bits of this, like private information retrieval, private set intersection, searchable symmetric encryption. All these sort of things give you some level of assurance that your data, your database provider can't, you know, accidentally read your data, or whatnot. Yeah. Service provider at all. I mean, like, all of this is tied together with trust, with contracts, with the ability to just promise that you're not going to do anything with the data. But these new privacy-preserving algorithms and these confidential computing mechanisms, they tie these things together in a way that's a lot more elegant. You don't have to trust as much. You can trust all you like, but this gives you something cryptographically verifiable, which, that's the magic.

Jon McLachlan

Yeah. One of the things that I've heard, someone's approach, they can [remain] nameless here for a moment, but they propose just to keep all of the data on the device, and make, since the devices are getting more powerful, faster, smaller, memory is cheaper, the silicon is cheaper, of like, you do more and more stuff on the device. And I thought about that, and I was like, well, yeah, but...

Nick Sullivan

Right? I mean, oh, yeah, [restoration]. Yeah. With this onslaught of new machine learning, LLM applications, data size is humongous. It's not something that fits on phones anymore. So you're going to have to have a hybrid architecture no matter what, if you're going to be doing anything with AI.

Jon McLachlan

Sure, they're going to be ginormous, and maybe a little bit slower, and the computer is going to be optimized for those machines up there in the cloud. There's all sorts of reasons why you would want to use a centralized SaaS service. And to be able to do that, like, and preserve your privacy at the same time, is ideal.

Nick Sullivan

And I'm hoping and expecting the industry to move in that direction. I don't know how far it'll go, or it'll succeed, but it's going to be a talking point.

Jon McLachlan

Is there anything that does not exist today that you just wish someone would sit down and build it already? A little bit of a leading question.

Nick Sullivan

This is the let's inspire your audience to come up with an idea segment.

Jon McLachlan

Yeah, it's a little bit of a leading question for all the [bored] entrepreneurs out there.

Nick Sullivan

Yeah, I don't have an idea to put on the table right now. There's so many ideas. And so, like, finding the, yeah, so a lot of different things that I ran into at Cloudflare, my colleagues have run into. These are a lot of the companies I'm advising, is like, hey, this whole thing is a huge pain, I'm going to just sit down and build something that solves it. And actually, a lot of the best ideas come from people who are just banging their head against the walls inside of a big company, trying to get something done that just requires, you know, a SaaS system or, like, some piece of software to [eliminate] all the manual work. And those are the types of companies that, some of the types of companies that inspire me, is like operators, operating companies, like, having generic problems that they know aren't being solved somewhere else, and just going out and solving it.

Jon McLachlan

Amazing. If you could go back and visit your younger self and share a piece of advice with your younger self, would you, and what would that advice be?

Nick Sullivan

This is such a hard question.

Jon McLachlan

Sorry for [putting you on the spot].

Nick Sullivan

It's fine, it's fine. No, but it's, it's good, it's a good one, right? I mean, like, it goes to the core of, like, yeah, mentoring yourself, or, like, what is your personal narrative, or, like, where did you get to where you are? Is it, like, butterfly effect, some sort of change in the past? And, and, you know, I'm relatively happy where I am, right? So it was, it was a good path. But I feel like I'm much more of someone who learns through experience and through making mistakes than from listening to someone, listening to some old guy, like, tell you what to do. Yeah. Somebody, somebody in their 40s telling you what to do as an early 20-something. But, and also, you'd barely have the context to be able to take advice to heart, unless you really think about it. And so I'd have to really think about it. But I know that, you know, as a young person, I was always in pursuit of novel ideas, and in pursuit of exciting things that are able to change the way that people interact, or people interact with technology. And I would probably tell myself to continue to lean into that, be curious, my young self, because curiosity unlocks, unlocks the world to you, right? And unlocks your ability to know yourself. So get out there. Follow your curiosity, I would say.

Jon McLachlan

Great advice. Nick Sullivan, everyone. Thank you so much for joining me on an episode of the Security Podcast in Silicon Valley, Nick.

Nick Sullivan

This is amazing. Thanks, Jon. And I really enjoyed this conversation, and good luck, and hope to see you online soon.

Jon McLachlan

Thanks for all of our listeners as well for tuning in. Thanks, everyone.

Nick Sullivan

Thanks, Jon.

FAQ

Questions this episode answers

Who is Nick Sullivan?
Nick Sullivan spent a decade at Cloudflare as head of security engineering, head of cryptography and the founder of Cloudflare Research, after six years on Apple's DRM team. He co-chairs the IRTF's Crypto Forum Research Group and joined the Internet Architecture Board in March 2026. He studied pure mathematics at the University of Waterloo and cryptography at the University of Calgary.
What does we make friends mean?
It was the slogan Nick gave Cloudflare's research team. Research has no short-term deliverables, so it depends on relationships and on people who want to work with it.
What is connection tampering?
Devices in the network path that block or disrupt connections, for reasons from corporate phishing protection and school content filters to censorship. A 2023 SIGCOMM paper with Nick among the authors measured it passively from Cloudflare's servers, using signatures in packet headers.
What is Privacy Pass?
A protocol that gives users anonymous tokens for solving a challenge, which they can spend later to skip challenges like CAPTCHAs. Nick was among the authors of the 2018 paper, and the IETF published it as RFCs 9576 to 9578 in 2024.
How does iCloud Private Relay work?
It routes Safari browsing, DNS and unencrypted app traffic through two relays. Apple runs the ingress relay, which sees your IP address but not your destination. A provider other than Apple runs the egress relay, which sees the destination but not your IP. Nick said Cloudflare took part.
What was the Cloudflare Heartbleed Challenge?
In April 2014, unsure whether Heartbleed could leak a server's private key, Cloudflare set up a vulnerable server and invited anyone to take its key. Outside researchers did within hours, and Wikipedia records at least two of them stealing it on April 12, 2014.
What is Messaging Layer Security?
An IETF protocol for end-to-end encrypted group messaging, published on July 19, 2023. It keeps properties Signal brought to two-person chats, such as forward secrecy and post-compromise security, with key agreement that scales to groups as large as 50,000. Nick co-chaired the working group when the episode aired.
Why does post-quantum cryptography matter now?
Someone can store encrypted traffic today and decrypt it once a quantum computer exists, an attack called harvest now, decrypt later. NIST finalized its first three post-quantum standards in August 2024, and Chrome turned on hybrid post-quantum key agreement by default in November 2024.
What are privacy-enhancing technologies?
Cryptography that provides some of what tracking did without tracking people. Nick's example was DAP, which uses multi-party computation to collect analytics privately.
What is confidential computing?
Protecting data while it's in use. Nick named homomorphic encryption, which computes on encrypted data, as the holy grail, along with private information retrieval, private set intersection and searchable symmetric encryption. They let you trust a provider less.
What advice does Nick Sullivan give his younger self?
Follow your curiosity. He learns through experience and mistakes, and curiosity, he said, opens up the world and helps you know yourself.
Keep reading on the YSecurity blog
New episodes every other Tuesday

Got a story like this one?

We're always looking for practitioners who've been in the room when it mattered.