Unsupervised Hierarchical Side-Channel Analysis on ML-KEM Cryptographic Algorithms

Post-quantum cryptography (PQC) is a new generation of cryptographic schemes designed to resist attacks from quantum computers on existing cryptographic algorithms. Among current PQC proposals, lattice-based cryptography has emerged as the most promising solution due to its simple structure, fast computation speed, and high security. Consequently, it has garnered significant attention from side-channel analysis researchers in recent years. Our work focuses on ML-KEM, and applies unsupervised learning techniques for its side-channel analysis. By collecting trace leakages during its inverse Number Theoretic Transform (NTT) operation, and using specific ciphertext traces for simple power analysis, we successfully recover the key using unsupervised learning algorithms. Compared to existing techniques, our proposed method can recover the key using a small number of traces.