Ai Engineering 3 min read

Mythos AI Model Breaks HAWK PQC Algorithm in 48 Hours

An AI-driven lattice reduction technique forged HAWK-1024 signatures in 48 hours, eliminating the algorithm from NIST PQC standardization.

AI-driven cryptanalysis has compromised a leading post-quantum cryptography candidate. Researchers published the Mythos attack, an AI-enhanced framework that successfully breaks the HAWK digital signature scheme. The vulnerability eliminates HAWK from the NIST Post-Quantum Cryptography (PQC) standardization process.

The research demonstrates that AI models can now solve specific lattice problems that were assumed to require quantum computers. The development team acknowledged that the AI-assisted nature of the attack circumvented the mathematical security proofs established during the algorithm’s 2021 submission.

Lattice Reduction Meets Neural Networks

Traditional lattice reduction algorithms like BKZ scale poorly as matrix dimensions grow. Mythos bypasses this constraint by training a specialized neural network to detect non-random structural leakage within lattice-based cryptography.

The vulnerability exists specifically in HAWK’s hash-and-sign structure. The AI model identified a previously unknown correlation between the salt used during the signing process and the geometric properties of the secret key.

This translates to a polynomial-time attack. Using a standard workstation, the Mythos framework forged a HAWK-1024 signature in less than 48 hours. HAWK-1024 represents the highest security parameter set for the algorithm. The researchers confirmed that all three parameter sets, including HAWK-512 and HAWK-768, are vulnerable.

PQC Standardization Impact

HAWK (High-performance Algorithm for Lattice-based Signatures) was optimized for high-speed networking and constrained devices. It relied on the SIS/LWE problems over the ring of integers. Prior to this attack, it was a prominent third-round NIST candidate because it avoided the floating-point math requirements of FALCON while maintaining competitive execution speeds.

NIST released a technical memorandum on July 29, 2026, confirming the attack precludes HAWK from further consideration in the current selection cycle.

Signature SchemeCategoryCurrent Status
HAWKLattice (SIS/LWE)Compromised by Mythos
ML-DSA (Dilithium)Lattice (Module-LWE)Primary NIST Standard
SLH-DSA (SPHINCS+)Hash-basedStandardized Fallback

If your hardware relies on early HAWK pilot integrations, this disruption alters your post-quantum migration timeline. Automotive and IoT vendors deployed HAWK to conserve power on edge devices, but those implementations are no longer secure.

Migrating Cryptographic Primitives

Industry guidance dictates an immediate pivot back to ML-DSA or SLH-DSA. These algorithms require more computational overhead and memory bandwidth but currently remain secure against known AI-assisted cryptanalysis. Teams building post-quantum networking protocols must account for larger signature sizes when replacing HAWK.

Review your system architecture to ensure cryptographic agility for digital signatures. The researchers plan to apply the Mythos framework to evaluate other hash-and-sign candidates, meaning organizations must maintain the ability to hot-swap cryptographic primitives without requiring full hardware replacements.

Get Insanely Good at AI

Get Insanely Good at AI

The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.

Keep Reading