

✓CTO at @ledger. Busy securing the blockchain revolution. Cryptography, (Hw) Security, Tech, Blockchain. Previously built the Donjon (@DonjonLedger)


Would be awesome to design a system where users don’t have to rely on third parties to own their value and can transact 24/7 without downtime...


🚨 Tor network under massive DDoS attack, emergency security update released as relays are overwhelmed The Tor Project has released an emergency security update, version 0.4.9.17, in response to an ongoing distributed denial-of-service (DDoS) attack affecting most relay operators across its network. Attackers are flooding relays with massive numbers of client circuits, overwhelming infrastructure and disrupting network operations. The update strengthens IPv6-based DoS protections by tracking clients by network prefix rather than individual IP addresses. Tor warns that defending against the attack remains challenging and additional security releases may be necessary. Relay operators are urged to upgrade immediately.

.@TechloreInc's founder Henry sat down with Ledger CTO @P3b7_ at @Ledger OP3N to talk about Ledger Atom: what it actually protects, where it stops, and what it means for privacy. Watch the full interview: youtube.com/watch?v=roTqSLSU…

Hash-based signatures are often pitched as the safe bet against quantum computers and AI. Ledger's CTO, Charles Guillemet, isn't convinced they deserve that status. The disagreement goes right to the heart of this debate: the experts can't agree on which kind of cryptography is safest. At Devcon this month, Ethereum Foundation researchers called hashes "the most conservative option" against quantum and AI attackers, and @VitalikButerin has advised favoring hash-based schemes wherever possible. Guillemet underscores that full SHA-256 is still secure, but AI is helping researchers eat into its safety margin. So hash-based signatures, as the Ledger CTO puts it, "are not fundamentally more immune to AI-based attacks than other signature schemes." Coinbase's head of cryptography, @LindellYehuda, has made a similar point from the history books: hash functions have been broken more often than elliptic curves. These are informed judgments. Nobody knows which kind of cryptography will still be standing in ten years. That's the strongest argument against a blockchain making one choice for all its users. A chain that writes one signature scheme into its rules is betting on the winner, and if the bet goes wrong, the fix is another fork. On CKB, nobody has to place that bet for everyone. A hash-based scheme, SPHINCS+, is live on mainnet today. If another scheme becomes the better choice, it can be added the same way, as software, and holders can move when they choose.

Congrats to the team on this one! Turning a prompt injection in Codex into a full sandbox escape + supply-chain PoC is genuinely impressive work. Looking forward to the next posts ;-)


Unfortunately this kind of vulnerability used to be very expensive and reserved to high potential attackers. Their cost is trending towards zero and are now exploitee at scale. Simply don't store any valuable in your mobile, especially not your crypto seeds...

Thank you 🙏 cumulatively we lowered round 39 by ~70 bits in a few weeks from literature Round 40 is next frontier and so far we haven't found a door

Great work on pushing the best known collision attack on 39-step SHA-256 below generic search. Full SHA-256 remains perfectly secure with a 25-step margin. This does show that AI-assisted cryptanalysis can meaningfully advance reduced-round attacks, which is relevant for assessing the long-term security margin of hash-based signatures. One more reason to think that hash-based signatures are not fundamentally more immune to AI-based attacks than other signature schemes.

More vulns found using AI since january than in the 8 years before from @DonjonLedger. Finding vulnerabilities is basically free now. Patching them remains costly, having strong security guarantees remains the challenge.

🤖OpenAI just published research showing models found and developed exploits for 948 local privilege escalation (LPE) vulnerabilities in the Linux kernel, spending $1.2M in API credits. To put this into perspective: the Linux kernel powers almost all cloud infrastructure, Android devices, supercomputers, and modern networking. Linux enforces process isolation. An unprivileged application, like a calculator or a basic utility, is not supposed to read memory, inspect processes, or touch cryptographic keys held by another process or software wallet. A local privilege escalation simply breaks that security boundary. It allows an unprivileged process to escape its world and gain root control over the entire system (like full access to everything). OpenAI demonstrated that an automated pipeline can discover nearly a thousand ways to do this in a single campaign. While the research is impressive, it exposes three uncomfortable realities: 1️⃣ Finding bugs is not fixing bugs LLMs are remarkably good at static analysis, reproduction, and exploit generation. Generating sound patches for complex kernel code is an entirely different challenge. Kernel fixes often introduce performance regressions, subtle race conditions, or fail to address the root architectural variant. 2️⃣ The attack surface remains massive These 948 exploits targeted only local privilege escalation, and only audited a fraction of the kernel code. They do not account for remote code execution, network stack vulnerabilities, firmware flaws, or supply chain compromises. Finding 1,000 paths to root does not mean the system is secure; it simply means we mapped 1,000 holes in a 40-million-line dam. The density of vulnerabilities in software is extremely higher than anyone thought before 🫠 3️⃣ The defensive economics do not scale A $1.2M audit run is pocket change for a frontier AI lab or a patient nation-state adversary. But virtually no open-source maintainer, startup, or enterprise can spend millions of dollars per release cycle on tokens just to partially reduce latent vulnerabilities. If defense costs millions of dollars per scan while sophisticated attackers only need to find one unpatched path, the asymmetry favors the attacker. The real lesson here is that software isolation alone will always fail under automated adversarial pressure. The only durable answer is architectural: minimizing attack surfaces, proving security invariant, enforcing strict hardware roots of trust where secrets live, and building toward cryptographic guarantees that make critical compromises structurally impossible rather than merely expensive.


I think the smileys represent Vitalik's mood. There's no consensus among cryptographers about that.

Most companies need to triple their cybersecurity budget for 2027, not 10% more. 3x more. At least. AI made recon and attacking basically free for threat actors. What you don't spend protecting, you will spend it 10x after an incident. If you are lucky to survive.