←
Federico Maggi5,172 posts
Federico Maggi avatar

Federico Maggi

@phretor

Cybersecurity engineer, researcher, and advisor with broad technical and scientific experience on designing, analyzing, and testing security-critical systems.

◷ Joined February 2008▧ 425 Photos and videos
Federico Maggi avatar
Federico Maggi@phretor·Nov 14

TIL that about 1 in 28 Linux kernel CVEs was introduced by an earlier CVE fix: 638 CVEs (3.6%), caused by 577 fixes (2.8% of fixes). This is a lower bound because we don't (yet) have all CVEs.

will depue avatar
⟳ Federico Maggi retweeted
will depue@willdepue·Nov 12

I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

Post attachment
CoreWeave avatar
⟳ Federico Maggi retweeted
CoreWeave@CoreWeave·Nov 12

Valvey and Racky, CoreWeave's rack systems for @nvidia Vera Rubin NVL72, are already in our data centers and in daily use. Alex Dutton says no one had ever put 200 kW in a rack before Vera Rubin. Bigger pumps and heat exchangers meant tighter control at every rack. Valvey and Racky handle that as part of CoreWeave Mission Control, and we're scaling them with our partners. Alex walks through it below. crwv.co/uuo0r

0:55
dawgyg avatar
⟳ Federico Maggi retweeted
dawgyg@thedawgyg·Nov 7

I'm winding down my time in the Apple Security Bounty program — so I'm open-sourcing the research. 100+ PoCs, real CVEs, and the fuzzers that found them. Public repo 👇 github.com/dawgyg/AppleResea…

Claudio Criscione avatar
⟳ Federico Maggi retweeted
Claudio Criscione@paradoxengine·Oct 30

Man the brotli thing really is way more widespread than I had realized. Lots of new detectors coming in the Tsunami scanner.

dawgyg avatar
⟳ Federico Maggi retweeted
dawgyg@thedawgyg·Oct 28

Since the @Apple Bug Bounty Program is by far the worst program I've ever hacked on, I am going to be dumping all of the data/POCs/report write ups etc for everything found during the 4 months of hacking them. This will include POCs for vulns that didnt meet their bar for a bounty so wasnt reported to them, so you may be able to find ways to escalate some of them. Ill be uploading the data all this weekend.

Federico Maggi avatar
Federico Maggi@phretor·Oct 27

[...] that make today's AI coding agents effective. I think vulnerability response lacks comparable infrastructure. We need to scale context, not repeatedly reconstruct it. Structured, reproducible, traceable vulnerability evidence should become a first-class artifact. [...]

Federico Maggi avatar
Federico Maggi@phretor·Oct 27

AI is scaling vulnerability discovery, but remediation still relies a lot on tribal knowledge and "human glue". I don't think LLMs will fix the missing infrastructure. The software industry & dev community spent decades building the foundations trustial.substack.com/p/repr… [...]

Burp Suite avatar
⟳ Federico Maggi retweeted
Burp Suite@Burp_Suite·Oct 21

🚨 PHP2Shell detection is now available in Burp Suite Professional & Burp Suite DAST. Most of the web is powered by PHP, which means most of the web is vulnerable to these critical unauthenticated remote code execution flaws (CVE-2026-109618 / 109619). You can already detect it in Burp Suite using: 🟧 ActiveScan++ v2.0.13 (portswigger.net/bappstore/31…) 🟧 Bambda scan checks (github.com/PortSwigger/bambd…) 🟧 BChecks (github.com/PortSwigger/bchec…) Start scanning for the vulnerability now → portswigger.net/bappstore/31…

Costin Raiu avatar
⟳ Federico Maggi retweeted
Costin Raiu@craiu·Oct 21

Currently seeing 196.189.236[.]67 probing for the PHP vuns, suggest to block them on gateways, been hitting honeypots with a bunch of stuff since at least Oct 2.

Federico Maggi avatar
Federico Maggi@phretor·Oct 22

The way #php2shell has been pre-announced is a glimpse of what the future of disclosure may be. Heads up to the public instead of embargoes, everyone rushes to find the same vuln, patches fly in, some will get exploited earlier. The best clankers will win the patching race.