Most companies still protect people on the web the way they did twenty years ago. Traffic leaves the laptop, passes through a proxy in the cloud, and the proxy decides from the address whether the page is safe. Vivek Ramachandran spent years helping red teams show their clients how easy that is to get around, and he came away convinced the decision has to be made somewhere else.
“because the rest of the industry is still looking at solving this in the cloud, while we are saying, solve this in the browser, run something browser-native entirely in the browser.”
Vivek Ramachandran, at 23:52
Vivek founded SquareX in May 2023 and ran it as founder and chief executive until Zscaler acquired it in February 2026. Before that he founded Pentester Academy, the online security training company INE acquired in 2021, and the Binary Security Group under which he built his training, products and services for more than a decade (LinkedIn). His DEF CON 32 speaker bio adds the rest: he discovered the Caffe Latte attack, broke WEP cloaking, built Wi-Fi monitoring and firewall tools, and in an earlier job was one of the programmers of the 802.1x protocol and port security on Cisco’s Catalyst 6500 switches (DEF CON).
Where we sit, so you can weigh what follows: YSecurity helps startups pass the security reviews enterprise buyers run, and how employees reach SaaS apps and the web comes up in almost every one of them. Jon McLachlan and Sasha Sinkevich co-hosted this one. We recorded in the summer of 2024, in the weeks before Vivek’s DEF CON 32 talk, and the episode aired on July 13, 2024.
Down to his knees
Jon opened with the simplest question on the show, how Vivek got into security, and the answer starts in 2000. Vivek was preparing for his engineering exams when a wave of denial-of-service attacks took down some of the biggest sites on the web. He heard about a hacker called Mixter, who had written the tool the attackers were said to be using.
“Now, I was barely 18 or 19, and what really intrigued me was, hey, how does one person have so much power that he’s actually bringing in all these massive websites down to his knees?”
Vivek Ramachandran, at 1:18
So he went looking for the code.
“And that was when I started researching, downloaded the source code of Trinoo, understood nothing about it, and then slowly decided, like, hey, there’s so much for me to learn.”
Vivek Ramachandran, at 1:29
The record backs the memory. Tribe Flood Network, the tool Vivek calls Tribal Flood Network on the tape, is a set of programs for distributed denial-of-service floods, and its 2K version was written by Mixter, a hacker based in Germany (Wikipedia). Trinoo is its cousin, and Wikipedia connects a trinoo network to the February 2000 attack on Yahoo (Wikipedia). Vivek peeled the onion from there, landed a few internships, and calls what followed a lifelong obsession.
Your relationship with power
Jon recognized the story. When he was a student at the University of Minnesota, he tried to explain to a professor why RSA and ElGamal fascinated him, and got a sharper answer than he expected.
“oh, so you’re into security because it changes your relationship with power.”
Jon McLachlan, at 2:15
Vivek added that these things tend to obsess you a lot more. You want to go out there and rule the world, he said, to look different and feel different. Then Jon turned to Sasha, whose story he had never heard on the show, and Sasha’s answer was about data rather than power.
“We as security professionals have the opportunity to uphold that statement.”
Sasha Sinkevich, at 3:05
The statement is that everyone has the right to privacy. Sasha came to security because data reveals so much about a person’s life, which turns out to be the thread of the whole conversation. Almost everything SquareX built is about seeing what happens on a page without shipping what’s on it somewhere else.
Lone warriors and team sports
Sasha asked how security has changed since the days when one person could take down a giant. Vivek started with the culture.
“I think 25 years back, I feel like you still had a lot of lone warriors building out tools, putting it out there.”
Vivek Ramachandran, at 3:37
He remembers his first DEF CON talk, in 2007, as a room of curious researchers trading notes. DEF CON’s archive lists it: “The Emperor Has No Cloak, WEP Cloaking Exposed,” given while he was a senior wireless security researcher at AirTight Networks (DEF CON). Back then, he said, you picked a research topic, published, and people found you. Now there’s a path of exams and certifications. And the job itself looked different.
“So if you ask somebody what is security, they would say, to me, security is learning how to configure a Check Point firewall or a Cisco PIX.”
Vivek Ramachandran, at 4:37
The attackers changed most. Nation states and organized crime sponsor hackers out of reach of any court, and defenders answered by banding together too.
“And that has actually made security more of a team sport, and not just like a solo lone wolf game.”
Vivek Ramachandran, at 5:13
He pointed to DEF CON itself, where most of the talks that year came from companies and researchers working together. Sasha added that it’s no accident security now has a red team and a blue team.
Red teams asked for a bypass
The idea for SquareX came out of Pentester Academy. Its customers included Fortune 500 companies and government agencies, and their red teams kept asking for the same thing: something new to show that the enterprise’s defenses could still be bypassed. Vivek went looking where people actually work.
“And that was the time I figured, even though the web browser is the most used enterprise application, if you think about it, people spend 95% of their time on the browser, maybe 5% in applications like Outlook, there was very little happening on the browser itself to secure it.”
Vivek Ramachandran, at 6:20
What protected people on the web was the proxy, under whatever name it was sold: secure web gateway, SSE, SASE. All of them look at network traffic to stop attacks that happen at the application layer, inside the page. That mismatch is what the red teams exploited.
“So when I was doing my research to help the red teams out, I figured that it was very easy to bypass these proxies, because they didn’t have application context.”
Vivek Ramachandran, at 6:54
A proxy sees requests. It doesn’t see which tab made them, which account is signed in, what the user clicked, or what the page assembled from the pieces it downloaded. Attackers can split something malicious into innocent-looking parts and rebuild it in the browser, after the proxy has waved each part through.
It has one function
Putting security inside the browser sounds obvious. Vivek’s point was that it isn’t easy.
“It’s not a software platform to distribute new products. It has one function, and to show you web pages.”
Vivek Ramachandran, at 7:29
SquareX spent its first stretch, he said, researching and collecting tribal knowledge about how to run a security product inside something never designed to host one. The payoff he claimed for getting it right was large.
“then this could fundamentally protect your enterprise employees almost 10x better than cloud proxies.”
Vivek Ramachandran, at 7:46
Ten times is his estimate, not a measurement, and it’s the thesis the rest of the episode defends with examples.
Six million and 100,000 users
Jon paused to put the company’s numbers on the record for listeners: more than 100,000 IT and security users within five months of launch, a $6 million seed round from Sequoia Capital Southeast Asia in May 2023, and a major enterprise release in May 2024. TechCrunch covered the seed when it was announced, describing a Singapore-based startup whose browser extension opened links and files in disposable browsers running in SquareX’s data centers (TechCrunch). Vivek announced it on X within hours.
Exciting News: SquareX (@getsquarex ) has raised USD 6M seed round from Sequoia Capital ( @SequoiaIndiaSEA)!
Vivek Ramachandran (@vivekramac) · May 10, 2023
TechCrunch’s account of why he built it matches the tape. At Pentester Academy, Vivek said, customers kept disabling security products that got in the way of their work, like a Word document from a contact flagged as malware, so he set out to build one that didn’t block people (TechCrunch). Jon added a note for the founders who listen.
“And things don’t change unless people take risks, right?”
Jon McLachlan, at 8:40
The URL is too broad
Sasha asked what a prospect wants solved when they first call. Vivek’s answer was the weakness he’d been exploiting for red teams. Proxies were invented a decade back, as he put it, and the main thing they filter on is an address.
“Now, unfortunately, in today’s world, the URL is too broad a security construct for you to just say this is whitelisted or this is blacklisted.”
Vivek Ramachandran, at 9:21
His example is one any engineering company will recognize.
“Lately, attackers have been zipping ransomware, encrypting it, which is password-protecting the zip file, and uploading it to github.com.”
Vivek Ramachandran, at 9:32
The proxy can’t open a password-protected archive, and it can’t block github.com, because the developers need it. So the file goes through on the strength of the domain it came from.
SquareX’s answer was to look at everything loading on the page through a zero trust lens, with the address as only one of the signals. That’s the reason Vivek kept the decision in the browser rather than at the network edge.
Same URL, two identities
The second example came from a prospect, and it’s quieter than malware but more common. People keep a personal Gmail and the company’s Google Workspace open side by side. When they upload a file, it can land in the wrong drive.
“A lot of times, when you try to upload a file, you might inadvertently end up uploading it to your personal Google Drive, because the URL is the same, drive.google.com.”
Vivek Ramachandran, at 10:40
A proxy sees the same address both times. The browser knows which account each tab is signed in to, so it can apply company rules to the company identity and leave the personal one alone.
“So, based upon that, we are able to slice and dice and apply policies to enterprise identity workflows versus your personal identity workflows.”
Vivek Ramachandran, at 11:20
This is a kind of data loss buyers worry about, because nobody meant to do it. For a startup preparing for a security review, it’s also a question the buyer’s team asks early: how do you keep company files out of personal accounts?
Dedicated browsers are a bad idea
Sasha asked the deployment question every buyer asks: is SquareX an extension or a whole new browser? Vivek didn’t hedge.
“I think dedicated browsers is a bad idea.”
Vivek Ramachandran, at 12:03
His evidence was Brave, a privacy-first browser that, the last time he checked, had only 40 to 50 million users. Asking every employee to switch browsers is a fight you don’t need to pick. Brave has grown since. It says it reached 100 million monthly active users and 42 million daily users in October 2025 (Wikipedia), so his figure sits closer to Brave’s daily count today. The point survives the correction. Even a browser with a loyal following took years to get there.
So SquareX ships as an extension on the browsers people already use, Chrome and Edge at the time, with Safari, Firefox and Opera on the way.
“Browser extensions have superpowers, in the sense that they can monitor browser events, page events, intercept, proxy, do a bunch of stuff,”
Vivek Ramachandran, at 12:41
Those powers cut both ways, which Vivek knew better than most. An extension that can watch every page can also be the attack. SquareX’s own research after the tape went straight at that, from extensions that bypass Chrome’s newer Manifest V3 rules to fake extensions that pass for password managers, covered in What happened after the tape.
Three ways to use AI
Sasha asked about AI, in a year when every pitch led with it. Vivek explained why his didn’t. The word is so keyword-stuffed, he said, it’s almost like Web3, or crypto, and people turn skeptical the moment they hear it. Then he got specific.
“How do we use AI? Well, three different ways.”
Vivek Ramachandran, at 13:54
The first is detection where the data lives. SquareX watches hundreds of thousands of events in the page, too many to ship anywhere, so the models run in the browser.
“And we have our own lightweight ML models, which are running in WebAssembly, which our browser extension ends up loading within the browser itself.”
Vivek Ramachandran, at 14:06
The second is the policy language. SquareX’s engine has hundreds of parameters, and Vivek was candid that a startup can’t ask administrators to take a certification course before they touch it.
“I think the greatest beauty of LLMs, in my humble view, is that now you have a language interface to interact with anything which is ultra-complex below it.”
Vivek Ramachandran, at 15:08
An administrator types the rule in plain language, and the product turns it into the parameters underneath. His example:
“hey, block all websites which are downloading encrypted zip files and which may be less than 90 days old.”
Vivek Ramachandran, at 15:52
The third is triage. Security, he said, is a needle in a haystack problem, so an AI copilot reads the events and summarizes them for the admin.
Privacy-safe by default
Sasha picked up the first point and pushed it further. Shipping page data to a server isn’t only impractical. It creates a second problem.
“And then you have to worry about the data privacy of that set of data that you have shipped back to the server.”
Sasha Sinkevich, at 14:30
Every byte that leaves the laptop becomes data the vendor has to protect, and a question in the customer’s security review. Vivek agreed and made it a principle.
“we are firm believers that if something can be done in a privacy-safe way, that should probably be the de facto way to do it, because it safeguards people’s data.”
Vivek Ramachandran, at 14:47
It also puts less on SquareX’s own servers. For any startup selling security tooling, that’s a sales argument as much as an engineering one. The less customer data you hold, the shorter your own security questionnaire gets.
A four-year course in a product
Sasha’s next point was about complexity. The industry has built very complex products for a very complex space, and Sasha gave Vivek credit for hiding that complexity.
“So props to you for trying to elevate, or rather hide, the complexity associated with some of the problems that functions in the space.”
Sasha Sinkevich, at 16:46
Vivek has watched administrators miss most of what their tools can do.
“So product feature discovery and feature use is a real problem.”
Vivek Ramachandran, at 17:13
Sasha described the usual onboarding, a lot of integration, a lot of dashboards, and what feels like a four-year course in the product. Vivek’s answer for a startup was blunt.
“No one cares. Make it easy, make it fast, and probably you have an opportunity to do something here.”
Vivek Ramachandran, at 17:43
Giving first
Jon asked about go-to-market, the part of a startup technical founders tend to skip. Vivek’s playbook came from Pentester Academy, which had built a very large community before it sold.
“Pentester, at the time of sale, had over 150,000 people on LinkedIn following us, 200,000 people on Twitter, 250,000 people on Facebook.”
Vivek Ramachandran, at 18:36
The rule he took from that is short.
“It is about giving first, rather than taking.”
Vivek Ramachandran, at 18:55
From 100,000 to 200,000
For SquareX, giving first meant releasing something useful before the enterprise product existed. Every security team gets suspicious links and files, and most open them in a virtual machine, buy expensive tools, or rely on isolation features that don’t show what’s happening.
“So what we did is, our isolation features, we put it out as a free extension.”
Vivek Ramachandran, at 19:46
The free extension did two jobs at once.
“That allowed us to test the stability of the extension.”
Vivek Ramachandran, at 19:58
The other job was the one he’d hoped for. Jon had cited 100,000 users. Vivek said they were now very close to 200,000, and security teams were writing in to ask for an enterprise version they could trial, because their people already used the free one.
The CISO writes the final check
Bottoms-up adoption gets a product into the building. It doesn’t get it bought. Vivek described the second prong.
“For inline security products, the CISO has to write the final check, underwrite the risk, if this is an early stage startup.”
Vivek Ramachandran, at 20:51
So he went to RSA and the large conferences, and built an advisory board that could open doors. On the tape he names Jeff Moss, the founder of DEF CON and Black Hat, as an investor and advisor, along with the head of Black Hat and founders in Silicon Valley who had sold their companies (Wikipedia). The point was warm introductions, so he could talk to CISOs about their problems instead of cold-pitching them.
RSA was an amazing experience for SquareX and I will do a series of posts in the coming weeks. This one is around teamwork - our singular goal was to get massive coverage during the event.
Vivek Ramachandran · May 12, 2024
Bleeding-edge research
The third prong is the one Vivek said he absolutely enjoys, and it’s the one that defined SquareX afterward.
“Not just writing white papers, which are a summary of what you want the world to believe, but rather pushing the boundary and edges of how attackers are breaking products,”
Vivek Ramachandran, at 21:38
The research had a date on it as we recorded.
“So this year, literally, we are doing a main stage talk at DEF CON on how attackers could trivially bypass secure web gateways and sneak through even well-known malware and malicious websites.”
Vivek Ramachandran, at 21:55
Jon summed up the shape of it.
“You command the expertise with thought leadership. You have bottom-up adoption of users, but you have top-down monetization for your company.”
Jon McLachlan, at 22:28
For a security startup, the third prong is also the hardest to fake. A buyer can tell a white paper from a working exploit. Original research earns a CISO’s attention and gives the sales team something to talk about that competitors can’t copy.
On every browser on planet Earth
Jon asked what success looks like. Vivek measures it by impact, and he declined to borrow Steve Jobs’s line about a dent in the universe.
“I always measure success of a company based upon the impact that it can have.”
Vivek Ramachandran, at 23:15
For SquareX, impact meant the industry coming around to its contrarian view, that attacks have to be caught where they happen. He didn’t want to talk about total addressable market. Everyone has a browser, so the opportunity is to be on all of them, for SquareX or for others who share the thesis.
“have an opportunity to literally reside on every browser on planet Earth.”
Vivek Ramachandran, at 24:42
That’s why the free version existed, he said, and why SquareX planned a free web antivirus for anyone in the community.
From EDR to BDR
Sasha connected the idea to access control. Most of the traffic that reaches company systems comes through a browser, so oversight of what’s happening in the page is a layer of assurance whether someone clicked the wrong link or had a session stolen.
“And one of the biggest threats to the entire system is through access control.”
Sasha Sinkevich, at 25:24
Vivek answered with history. Fifteen or twenty years ago, work happened in desktop applications, and endpoint security grew up to protect them, from antivirus to endpoint detection and response and then extended detection and response. Now work happens in websites.
“Almost 15, 20 years back, I think all work was happening on the desktop.”
Vivek Ramachandran, at 26:13
He expects the browser to go through the same evolution, and he named the category.
“Something I like to call BDR, a stealing from EDR.”
Vivek Ramachandran, at 26:50
Browser detection and response became SquareX’s category name. Its site still calls the product the industry’s first (SquareX), and Zscaler’s announcement frames the acquisition as extending zero trust into the browser (Zscaler).
Three kinds of CISO
Jon asked for the pitch. Is it certifications like ISO 27001 and SOC 2, or a conversation about real security? Vivek said it depends on who’s across the table, and he sorts CISOs into three buckets before the call or in its first few minutes.
“I try to understand what kind of CISO is this gentleman or lady. And I kind of put them in three buckets.”
Vivek Ramachandran, at 27:34
The first are hyper-technical, former researchers and engineers who interrupt the moment he says browser-native.
“There, I’m a quiet listener, and I get bombarded with super technical questions.”
Vivek Ramachandran, at 27:57
The second came up through risk and compliance and lean on analyst reports and vendor comparison charts. A new category doesn’t show up on those charts yet, so he asks about the gaps instead.
“hey, what are the use cases none of your vendors have solved, especially the web proxies that you use?”
Vivek Ramachandran, at 28:29
The third have already been burned. They walk in with the attack that hit them, a single sign-on identity hijacking attack in his example, and ask whether you stop it.
“And I think they want literally a pointed response, yes or no.”
Vivek Ramachandran, at 29:24
Sasha’s aside fits founders in any category: when you create something new, there’s no concept of it in the buyer’s head yet. Use cases do the work a category name can’t.
A clean link to a fake login
Sasha asked for a public example, an attack SquareX could have stopped. Vivek picked the most common attack organizations worry about, spear phishing, and explained why it got harder to spot.
“And now what’s been happening is, with LLMs, the mails sound better, they’re more contextual, the websites look better, right?”
Vivek Ramachandran, at 30:04
The version he described starts with a link to SharePoint, on a private tenant the attacker created. The page asks you to fill in a form or download a document. Nobody’s proxy blocks SharePoint.
“Now, SharePoint, literally every web proxy is going to whitelist by default, right?”
Vivek Ramachandran, at 30:36
The SharePoint page opens a third-party site, the third-party site shows what looks like a Microsoft login, and because the journey started on a real Microsoft address, people sign in again without thinking about it.
“And that’s really where the keys to the kingdom are lost.”
Vivek Ramachandran, at 31:25
The proxy saw every step and couldn’t connect them.
“because web proxies are only looking at individual URL requests in a stateless way.”
Vivek Ramachandran, at 31:38
Read-only, request an exception
The company that brought SquareX this attack wasn’t sure it could be solved. Inside the browser, SquareX could see the chain.
“And incredibly, because we sit in the browser, we can actually create a hierarchical tree,”
Vivek Ramachandran, at 32:00
The tree shows that a SharePoint tab opened a site outside the company’s known domains, and that the site is now asking for input. The fix was a policy rather than a detection. Pages opened from a SharePoint tenant that isn’t on the allowlist can be visited, but nothing can be typed into them, passwords included. Sasha named the pattern.
“And you are essentially a pre-hook into that logic. You block the action before it’s activated.”
Sasha Sinkevich, at 32:40
When someone tries to type, the field is blocked and the page explains why.
“we end up blocking and saying, this is read-only, request for an exception.”
Vivek Ramachandran, at 33:03
It’s a good model for any control a startup designs, browser or not. Don’t try to recognize every fake page. Use what you already know about how the user got there, and make the risky step need a reason.
Faces at RSA
Jon asked for the best day so far. Vivek picked RSA Conference 2024, where SquareX showed its enterprise product and people came up having already used the free extension. One visitor stood out, under NDA, so Vivek didn’t name the company.
“I personally use your disposable browsers and file viewers to detonate suspicious documents and links.”
Vivek Ramachandran, at 34:23
That was the head of cybersecurity at a Fortune 50 company, who said the tool was now being taught to the IT team too. SquareX’s recap video from the conference is still up on its channel (YouTube).
Funding is responsibility
Vivek used the same answer to say what doesn’t count as a best day.
“I personally don’t think funding is a milestone to be super happy about. It’s nothing more than more responsibility.”
Vivek Ramachandran, at 34:45
Jon put it in founder terms.
“You’re selling a piece of the company for a fair value, and you’re giving up a little bit of control for a little bit of cash.”
Jon McLachlan, at 34:57
Hiring didn’t make Vivek’s list either, for the same reason. What he tells everyone at the company is simpler.
“measure an impact. That’s it. Has your work made impact?”
Vivek Ramachandran, at 35:17
The RSA day counted because it was the first time people told him so in person. Sasha suggested a bigger version of the same test.
“Today, if we want to search for something, we say, Google it.”
Sasha Sinkevich, at 35:54
The hope was that one day people would say the same about protecting their privacy in a browser, with SquareX as the verb.
The first ten minutes
Jon asked for the hardest day, and Vivek told a story without a happy ending. Through a warm introduction he met the CISO of one of the largest consumer electronics companies in the world. The demo went well, and the CISO asked to try it himself. SquareX set up a new tenant for the trial, and by mistake it shipped with experimental policies the team was still testing.
“Within the first seven or 10 minutes, some of those experimental policies ended up hitting, and there were some false positives, and he uninstalled it.”
Vivek Ramachandran, at 37:42
Jon’s aside will make every startup engineer wince.
“And doing great if those are actually separate environments, too.”
Jon McLachlan, at 37:20
They were, Vivek said. The trouble was in what got deployed where. The team rolled back, Vivek wrote to the CISO, and he heard nothing back. Eventually he ran into the CISO at a mixer, and the CISO explained.
“You need to make sure that you understand that someone like myself, if I give you time and if I give you opportunity, that is a privilege.”
Vivek Ramachandran, at 38:10
The next chance would be a year away.
“So come back to me exactly in a year.”
Vivek Ramachandran, at 38:31
What SquareX changed was its onboarding. Several people now check every new tenant, and Vivek does the final check himself, checking all the knobs before anyone outside sees it. He didn’t dress it up.
“Sometimes there is no comeback, and there’s no sweet story at the very end.”
Vivek Ramachandran, at 39:04
For a security startup, the lesson is concrete. A trial tenant is production for the person trying it. Treat its configuration like a release, with a checklist and a second pair of eyes.
Not a failure unless you give up
Sasha offered a saying for founders.
“If you’re not making mistakes, you’re not doing enough.”
Sasha Sinkevich, at 39:14
Jon added the other half.
“not a failure unless you give up.”
Jon McLachlan, at 39:40
Vivek’s practice after a mistake is about the team. Nobody plays the blame game, because the people who made the mistake already feel as bad as anyone.
“It’s very important as a leader to be supportive through mistakes, rather than purely point out mistakes.”
Vivek Ramachandran, at 39:52
He talks through the fix, tells everyone it’s fine, and takes the team to lunch somewhere with good desserts. Jon’s verdict was short.
“It’s just software, right? Life goes on.”
Jon McLachlan, at 40:24
Start earlier, then be patient
Jon asked what Vivek would tell his younger self. The first answer was to start sooner.
“I would probably have told myself, start even earlier.”
Vivek Ramachandran, at 40:57
He explained why he hadn’t. Growing up in a middle-class family in India, with parents working hard to give him the best education they could, he absorbed a fear of not having a job, and it made him risk-averse. His parents supported him, he said, and still asked whether he should work a few more years first. The second answer was the harder one.
“I have now become the firmest believer in the power of slow compounding.”
Vivek Ramachandran, at 41:48
Compounding asks for patience, and patience was what his younger self lacked. Now, when a proof of concept takes four months to start and ten meetings with different people, he reminds himself what it is.
“hey, this is the process. So enjoy the process.”
Vivek Ramachandran, at 42:39
Celebrate small wins
Vivek also disagreed with a piece of startup folklore, that nothing counts until the company is worth a billion dollars.
“My perspective is to celebrate every small win.”
Vivek Ramachandran, at 43:25
He doesn’t think it makes anyone complacent. Small wins give the team proof it’s moving, and that compounds too.
“Make sure your team feels that they are winning as well.”
Vivek Ramachandran, at 43:28
A force multiplier for go-to-market
Jon closed with a question for the founders listening: what tool do you wish someone would build? Vivek’s answer was for technical founders like him.
“GTM always is a blind side, right?”
Vivek Ramachandran, at 44:44
Founders who love building assume people will come. What he wanted was a way to make one piece of work go further.
“how can you take a piece of original content and then create different versions of it, which can be disseminated in an automated way across multiple platforms?”
Vivek Ramachandran, at 45:04
Take one video demo, he said, and turn it into a PDF report, a narrated podcast and other collateral, then publish it everywhere automatically. He was sure language models would get there, and admitted SquareX had tried to build it internally and hadn’t yet.
It’s the same giving-first instinct he described earlier, applied to content: one piece of real work, offered in every format people read.
What happened after the tape
A few weeks after we recorded, Vivek and SquareX’s chief architect Jeswin Mathai gave the talk he’d promised, at DEF CON 32 on Friday, August 9, 2024. Its title was “Breaking Secure Web Gateways (SWG) for Fun and Profit,” and the abstract introduced a new class of attacks, Last Mile Reassembly, that the authors said could bypass every secure web gateway in the Gartner Magic Quadrant for SASE and SSE as of its writing (DEF CON). SquareX released a site where anyone could test the bypasses against their own gateway (browser.security), and its recap counted more than 30 methods (SquareX Labs).
Two weeks later Vivek explained the research on Enterprise Security Weekly, in a segment SquareX sponsored.
The research kept coming, and much of it went after the same blind spot from the tape, what happens inside the browser where nothing upstream can see it. In October 2024 SquareX showed extensions that still get past Chrome’s Manifest V3 rules. In December it warned about consent phishing aimed at extension developers, the technique behind the Cyberhaven breach, in which attackers took over Chrome Web Store accounts to push malicious updates (SquareX). Vivek’s post that week walks through it.
Can a Chrome Extension be taken over from the Chrome Store with just a few clicks? SquareX has uncovered targeted attacks on Chrome Extension developers aimed at taking over the chrome extension from the Chrome store
Vivek Ramachandran · December 21, 2024
In 2025 SquareX ran what it called the Year of Browser Bugs, a commitment to publish architectural vulnerabilities in the browser, modeled on the month-of-bugs projects of the 2000s. Its research page lists 11 pieces, including browser syncjacking in January, polymorphic extensions that impersonate password managers in February, browser-native ransomware in March, data splicing attacks on data loss prevention tools at BSides San Francisco in April, and Passkeys Pwned, a flaw in how passkey registration and sign-in can be abused, in August (SquareX). The later pieces turned to AI agents and AI browsers, the newest software to sit between a person and the web.
Then, on February 5, 2026, Zscaler announced that it had acquired SquareX and closed the deal the same day, with terms undisclosed. The release says SquareX will let organizations secure users inside the browsers they already use, Chrome and Edge, without a full agent or a separate third-party browser (Zscaler). That is the argument from Dedicated browsers are a bad idea, now in Zscaler’s own words. Vivek’s note on X that week thanked the investors who had backed the seed.
Some great news! Zscaler @zscaler has acquired SquareX @getsquarex this week
Vivek Ramachandran (@vivekramac) · February 8, 2026
Eight months later he described what came next inside Zscaler.
The past few months at Zscaler since the SquareX acquisition have been nothing short of exhilarating! We brought in our technology into Zscaler's Zero Trust Browser, which is now being deployed with some of Zscaler’s largest customers.
Vivek Ramachandran · October 14, 2026
The thesis from the tape held up. In 2024 the bet was that security had to move into the browser, through an extension rather than a new browser, and that research would make the case. By 2026 the browser had become a category the big platforms buy into, and one of them bought SquareX. On episode 98, Or Eshed of LayerX told the same story from another company in the category, which Akamai bought in 2026. For a startup, the takeaway is closer to home. If your people work in a browser, your security review will ask what happens there, and “we have a proxy” is only part of the answer.
Did you catch it? A six-question quiz
Interactive · six questions
Did you catch it?
Six questions on what Vivek, Jon and Sasha actually said, with the timestamp for each answer. The answers are under each question for anyone reading without JavaScript.
-
Which tool's source code did Vivek download as a teenager?
- A port scanner
- Trinoo, a denial-of-service tool
- An early web browser
b. "And that was when I started researching, downloaded the source code of Trinoo, understood nothing about it" Transcript, 1:29.
-
What does Vivek call "too broad a security construct"?
- A password
- An IP address
- The URL
c. "Now, unfortunately, in today's world, the URL is too broad a security construct" Transcript, 9:21.
-
Why did SquareX ship as an extension instead of its own browser?
- Getting everyone to switch browsers is a losing fight
- Browsers can't run security code
- Extensions are cheaper to sell
a. "I think dedicated browsers is a bad idea." Transcript, 12:03.
-
Where do SquareX's lightweight detection models run?
- In a cloud proxy
- In WebAssembly, inside the browser
- On the email server
b. "And we have our own lightweight ML models, which are running in WebAssembly" Transcript, 14:06.
-
What did SquareX's policy do on pages opened from an unknown SharePoint tenant?
- Blocked SharePoint entirely
- Sent the page to a sandbox
- Made them read-only, with a way to request an exception
c. "we end up blocking and saying, this is read-only, request for an exception." Transcript, 33:03.
-
After the trial that failed, what did the CISO tell Vivek to do?
- Come back in exactly a year
- Send a new build next week
- Talk to procurement
a. "So come back to me exactly in a year." Transcript, 38:31.
What does your proxy miss?
Vivek’s test for any web control is what it can see. This scorecard asks eight questions about where your people work online, from personal accounts to extensions to phishing chains. It takes about two minutes.
Scorecard · eight questions
What does your proxy miss?
Answer for the laptops and browsers your people use for company work, including contractors. If two answers fit, pick the lower one.
-
Can your controls tell a company Google Drive or OneDrive from a personal one at the same address?
- No, both are the same site to us
- For some apps, through tenant or account restrictions
- Yes, policy follows the signed-in identity in each tab
"because the URL is the same, drive.google.com." See Same URL, two identities.
-
What happens when someone opens a link from an unfamiliar SharePoint or OneDrive tenant and is asked to sign in again?
- Nothing stops them typing a password
- They see a warning
- Pages reached that way are read-only, with an exception path
"we end up blocking and saying, this is read-only, request for an exception." See Read-only, request an exception.
-
What happens to a password-protected archive downloaded from a site you trust, like github.com?
- It's allowed, because the site is
- It's flagged for review
- It's opened somewhere disposable, or blocked by policy
"Lately, attackers have been zipping ransomware, encrypting it, which is password-protecting the zip file, and uploading it to github.com." See The URL is too broad.
-
Do you know every browser extension installed on company browsers, and what each can read?
- No
- We have a list
- An allowlist, enforced, with a review when an extension updates or asks for new permissions
Extensions "have superpowers," in Vivek's words, so an unreviewed one can read what your people read. See Dedicated browsers are a bad idea.
-
Where do people open suspicious links and attachments?
- In their own browser, and hope
- In a virtual machine, when someone remembers
- In a disposable browser or file viewer, by default
"I personally use your disposable browsers and file viewers to detonate suspicious documents and links." See Faces at RSA.
-
Can your web controls see content the page assembles after it arrives, through WebAssembly, file chunks, or WebSocket and WebRTC channels?
- They decide on the address
- They inspect some downloads
- Inspection happens in the browser, after the page is built
"Attackers could reassemble, do attacks on the browser." See Red teams asked for a bypass.
-
How do you know your web controls work?
- We trust the vendor
- A yearly penetration test
- We test known bypasses against our own setup, regularly
SquareX's test site lets anyone try the DEF CON 32 bypasses. See What happened after the tape.
-
How easy is it for your admins to change a browser security policy?
- Only one person knows how
- A few people, with documentation
- Anyone on the team can describe the rule and review the result
"So product feature discovery and feature use is a real problem." See A four-year course in a product.
Your web protection decides on the URL, which leaves personal accounts, trusted domains and phishing chains to the people clicking. Start with the two cheapest wins: separate company and personal identities in the browser, and make pages reached from unfamiliar file-sharing links read-only.
You see part of what happens in the page. Close the gaps attackers reach next: unreviewed extension updates, archives from trusted sites, and content assembled after it arrives. Then test the bypasses against your own setup.
Policy follows the user and the page, not just the address. Keep it that way as AI browsers and agents start doing the clicking: they inherit the same sessions and should cross the same checks.
How much of the work is in the browser?
Vivek’s estimate is that people spend 95% of their working time in a browser. Put in your own team to see how many hours that is, and how few happen anywhere else.
Calculator
How much of the work is in the browser?
Three estimates for your team: how many people, how many hours they work in a week, and the share of that time spent in a browser. Vivek's estimate on the tape is 95%.
A team of 50 working 40 hours a week spends about 1,900 hours a week in the browser, 7.6 hours a day each, or 91,200 hours a year. That's where the sign-ins, the uploads and the clicks happen, so that's where the checks need to be.
95% is Vivek's estimate on the tape, not a measurement. Use your own numbers.
From WEP cracks to Zscaler
Everything on this page sits on one timeline, from the denial-of-service attacks of 2000 to the Zscaler acquisition. Filter it by thread.
Timeline · 2000 to 2026
From WEP cracks to Zscaler
Filter by thread. Dates are from the public record except where marked as part of this show.
- The attacks that started it
A wave of distributed denial-of-service attacks takes down Yahoo and other large sites. Wikipedia connects a trinoo network to the Yahoo attack (Wikipedia).
- Engineer
Vivek starts as a developer, architect and consultant at companies including Cisco and AirTight Networks (LinkedIn).
- DEF CON 15
His first DEF CON talk, The Emperor Has No Cloak, on WEP cloaking, as a senior wireless security researcher at AirTight Networks (DEF CON).
- Binary Security Group
Founds the parent group for his security education, products and services (LinkedIn).
- Pentester Academy
Founds an online, subscription security training company, with offices in the US, Singapore and India (LinkedIn).
- Microsoft Regional Director
Microsoft names him a Regional Director of Cybersecurity for three years (DEF CON).
- Acquired by INE
Pentester Academy joins INE, and its cloud lab infrastructure moves into the INE platform (LinkedIn).
- SquareX and a seed round
Founded in May, with $6 million from Sequoia Capital Southeast Asia for disposable browsers delivered through an extension (TechCrunch).
- Email attachments
SquareX research reports that top email providers failed to detect and block malicious attachments (SquareX).
- SquareX Enterprise
Unveiled at RSA Conference 2024, the day Vivek calls his best at the company (YouTube).
- Black Hat Arsenal
Joins the Black Hat Arsenal Review Board (DEF CON).
- We record episode 49
"solve this in the browser, run something browser-native entirely in the browser." (Transcript).
- Episode 49 airs
On Podbean, Apple Podcasts and Spotify (Apple Podcasts).
- DEF CON 32
Vivek and Jeswin Mathai present Breaking Secure Web Gateways (SWG) for Fun and Profit, and the Last Mile Reassembly attacks (DEF CON).
- Sneaky extensions
Malicious extensions that still get past Chrome's Manifest V3 controls (SquareX).
- Consent phishing
Targeted attacks on extension developers to take over their Chrome Web Store accounts (LinkedIn).
- The Year of Browser Bugs
A year of published browser research begins with browser syncjacking (SquareX).
- Polymorphic extensions
Extensions that impersonate password managers and crypto wallets (SquareX).
- Browser-native ransomware
Ransomware that works in the browser rather than on the endpoint (SquareX).
- Data splicing
First shown at BSides San Francisco, techniques that slip data past data loss prevention tools (SquareX).
- Passkeys Pwned
An attack on how passkey registration and sign-in are implemented (SquareX).
- Brave at 100 million
The privacy-first browser Vivek used as his example reports 100 million monthly users (Wikipedia).
- Zscaler acquires SquareX
To extend zero trust into standard browsers without a third-party browser (Zscaler).
- Thanks to the seed investors
Vivek remembers the live browser attack demos in the seed pitch (X).
- Episode 98
Or Eshed of LayerX on browser security from another startup in the category (episode 98).
- Inside Zscaler
SquareX technology in Zscaler's Zero Trust Browser, and a new Endpoint AI Security product (LinkedIn).
What to do this week
Five things, in the order Vivek’s argument suggests.
- Separate the identities. Find out whether anyone can upload company files to a personal Google Drive or OneDrive from a company laptop. If they can, restrict sign-in to company accounts on managed browsers, or add a browser control that applies policy per identity. The scorecard above is a quick first pass.
- Inventory your extensions. List every browser extension on company browsers and what it can read. Move to an allowlist, and review an extension again when it updates or asks for new permissions.
- Make one risky step need a reason. Pick the riskiest action in your browser workflow, like typing a password on a page reached from an unfamiliar file-sharing link, and make it read-only with an exception path.
- Give suspicious files somewhere safe to go. Decide where people open the links and attachments they aren’t sure about, and make the safe place the easy one.
- Test your own web controls. Try known bypasses against your setup, or have someone do it for you. If you’d like a second pair of eyes, our penetration testing and product security teams do this work every week, and you can book 15 minutes with us.