Nadir (@kapytein) header
Nadir (@kapytein) profile picture
The Netherlands
Joined February 2018
Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)Photo by Nadir (@kapytein)
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
Nadir retweeted
The AI Agent Security - Multi-Step Tool Attacks simulation is now live! In partnership with @OpenAI, @Google and @IEEEorg, your challenge is to build an attack algorithm that stress-tests tool-using AI agents in a deterministic offline benchmark.
6
23
5
89
10,567
No, we explicitly mention that same-tenant is out of scope for some scenarios. Cross tenant has always been in scope.
1
1
1
26
We are revamping the bug bounty program's scope soon though, so keep an eye out for any updates.
1
1
30
Not having it in scope doesn’t mean we don't care about it perse! The focus of the BBP was on different vuln. classes, and e.g. internal efforts could get in the way of researchers, so having a vuln. class (temp.) out of scope actually helps to avoid that.
1
1
25
This reverse engineering work led its way to a fix in Chrome (subsequently Electron), with credit given back to me! Very cool to see, and happy to help the macOS ecosystem. I hope macOS fixes this huge issue soon. chromium-review.googlesource…
Everything is open source if you try hard enough. (Trying to find the source of a pathological performance issue stemming from AppKit only on macOS 26. I'm pretty sure it's a macOS 26 bug but given this is shipping, I need to find a workaround). I really wish Apple would just make the source of their frameworks available (even under a non-OSS license that doesn't allow any reuse), just so that app developers can understand how certain logic interacts with the system.
15
30
609
79,325
when building ai products, the key is having some reliable signal about what is working and what is not. whether that signal comes from evals, user feedback, your own taste, performance metrics, or just gut feel does not really matter as long as it is giving you actionable information about where to focus your efforts. the mistake is either having no signal at all (just building blindly) or getting caught up in the methodology of the signal rather than its utility. a formal eval that tells you nothing useful is worse than informal feedback that clearly points to real problems
i did not expect to wake up this morning and write a blog post
1
2
31
7,026
Replying to @rez0__
Bug bounty hunting, on at least web applications, has always been about discovery instead of exploitation. Exploitation has always been straightforward in most cases (excluding escalating issues into more severe bug classes). So, the work here by @Xbow is definitely impressive.
112
A fix from Google was released today. Part of the issue was due to my misunderstanding based on previous reports. Big thanks to chromium team for the quick resolve I hope everyone had some fun, and apologies to the triagers on HackerOne XD
Today I used a technique that’s probably not widely known in the community. In what cases could code like this lead to a vulnerability? ->
13
26
4
199
43,712
Imagine not having had any food to give your kids for 10 days.
I swear to God, this what exactly happened: I woke up today morning and found a missed call on my phone coming from a widow with a number of orphan children living next door. I called her back and asked her what she wanted. She said: “I need some flour. Me and my children have been sleeping without food for 10 days.” I replied: “Unfortunately, me and my children have nothing to eat for more than 10 days.”
229
3,041
38
11,017
498,757
Today I used a technique that’s probably not widely known in the community. In what cases could code like this lead to a vulnerability? ->
20
189
18
1,042
277,152
Replying to @Rhynorater
Mostly because CSPT are not 1:1 reproducible across targets. It’s a novel approach every single time (esp. when chaining bugs), which makes it a beautiful bug class.
1
1
397
We did it! We tested 300 Bay Area foods for plastic chemicals. We found some interesting surprises. Top 5 findings in our test results: 1. Our tests found plastic chemicals in 86% of all foods, with phthalates in 73% of the tested products and bisphenols in 22%. It's everywhere. 2. We detected phthalates in most baby foods and prenatal vitamins. 3. Hot foods which spend 45 minutes in takeout containers have 34% higher levels of plastic chemicals than the same dishes tested directly from the restaurant. 4. The 1950s Army rations we tested contained surprisingly high levels of plastic chemicals. 5. Almost every single one of the foods we tested are within both US FDA and EU EFSA regulations. Check out our full results below.
I'm going to re-run all these tests on food we eat in California. Also going to test for other plastic chemicals. Let me know what foods we should test and suggestions for methodology.
558
2,712
726
15,068
9,777,008
Replying to @intigriti
All the best in your next adventure! 🙌🏽
1
127
Nadir retweeted
in a couple weeks, i built a nuclear fusor in my bedroom – with zero hardware experience the secret? Claude sonnet 3.5 + projects a glimpse into the process below
314
992
567
9,798
3,193,436
Pass-the-{token} attacks are still very much relevant. Tokens may change: Cookie, NT Hash, Kerberos ticket, MFA token, ... However, the problem is not in the "token" but in the "pass". We need a solutions to make tokens stay put, such as device and channel binding.
Microsoft has detected a 111% year-over-year increase in token replay attacks, and incidents are continuing to grow. msft.it/6011lSgZ7
5
40
149
21,654
Replying to @charafmrah @benawad
IMO it is fine. There is usually other ways to achieve persistence, e.g. by inviting your own user account to an affected tenant, changing the user's e-mail address, linking a social login account, creating an API key, etc. The impact of a XSS doesn't really change.
1
426
Nadir retweeted
My latest blog about my discovery for Evernote Client All-platform RCE via PDF.js font-injection to preload.js exposed ipcRenderer-BrokerBridge-boron.actions bypassing Electron's nodeIntegration | context-isolation; Enjoy reading! 0reg.dev/blog/evernote-rce
10
122
2
458
33,048
happy to release my new article entitled: Next.js and cache poisoning: a quest for the black hole zhero-web-sec.github.io/rese… good reading;
38
179
3
770
84,560