AI Model Uncovers Vulnerabilities in Post-Quantum Cryptography
An advanced artificial intelligence model has identified previously unknown weaknesses in cryptographic algorithms, including one designed to resist quantum computing. Anthropic, the AI company, announced that an unreleased version of its powerful model, Claude Mythos Preview, discovered two novel attacks.
One of the identified attacks targets HAWK, a digital signature system intended to verify transaction authenticity without revealing private keys. HAWK was specifically engineered to withstand the threat of quantum computers and had advanced to the third round of the U.S. National Institute of Standards and Technology’s (NIST) post-quantum competition in May. It was the sole remaining lattice-based candidate in that stage.
Claude Mythos reportedly found a mathematical symmetry within HAWK that had eluded human researchers. This discovery significantly reduces the computational cost of compromising HAWK’s smallest key, from an estimated 2^64 operations to 2^38 operations – a reduction of approximately 67 million times.
Implementing a fix for HAWK would necessitate roughly doubling its key sizes. Anthropic noted that this change “eliminates many of the reasons” HAWK was initially appealing. In the context of blockchain technology, key size directly impacts block space, which in turn affects transaction fees. Therefore, any blockchain considering a quantum-resistant replacement would need to weigh the trade-off between signature size and security. HAWK’s primary selling points were its compact keys and rapid signing capabilities, advantages that would be diminished by the proposed fix.
For cryptocurrency holders, it is important to note that HAWK has not been implemented in any live systems. Bitcoin, for instance, continues to utilize ECDSA, the pre-quantum cryptographic scheme that these new candidates are intended to eventually replace. Consequently, no existing cryptocurrency infrastructure has been compromised.
However, the broader implication lies in the ongoing uncertainty surrounding post-quantum security, which remains a significant concern for the cryptocurrency industry. The extensive efforts underway to develop quantum-resistant solutions – including initiatives by entities like BlackRock and Coinbase, as well as roadmaps from Vitalik Buterin, NEAR, and Zcash – are predicated on the assumption that the proposed post-quantum cryptographic schemes are robust. The discovery by Claude Mythos, which identified a flaw in a leading candidate before its official release, raises questions about the soundness of these future security measures.
On a positive note, the same AI technology that can identify vulnerabilities can also be used to strengthen defenses. This is precisely why the industry is actively employing AI to scrutinize its own code. The current challenge is to determine whether AI can help rebuild cryptographic defenses more effectively than it can find ways to break them.
Leave a Reply