wolfSSL's AVX2-optimized ML-KEM implementation (mlkem_cmp_avx2) compares only 1536 of the 1568 ciphertext bytes during the Fujisaki-Okamoto re-encryption check in ML-KEM-1024 decapsulation. Ciphertexts that differ from the expected re-encryption solely in bytes 1536-1567 bypass implicit rejection and are accepted as valid, breaking IND-CCA2 security. An attacker able to submit chosen ciphertexts to a decapsulation oracle that uses a static ML-KEM-1024 key, and to observe whether the genuine shared secret or the implicit-rejection secret was produced, can use this as a plaintext-checking oracle to recover the private key. A proof of concept recovered a full ML-KEM-1024 private key with approximately 98% success using roughly 350 chosen ciphertexts. The flaw is a deterministic logic error and does not rely on timing measurements.

ML-KEM-1024 x64 AVX2 implicit rejection failure in the Fujisaki-Okamoto transform breaks IND-CCA2 security, allowing decapsulation to deviate from the implicit-rejection behavior required by the standard. The AVX2 constant-time ciphertext comparison used during decapsulation never compared the final 32-byte block of the 1568-byte ML-KEM-1024 ciphertext, so a ciphertext manipulated only in those final bytes would compare as equal and decapsulation returned the real shared secret instead of performing the required implicit rejection.

Affected: Ubuntu:16.04:LTS: wolfssl, Ubuntu:18.04:LTS: wolfssl, Ubuntu:20.04:LTS: wolfssl, Ubuntu:22.04:LTS: wolfssl, Ubuntu:24.04:LTS: wolfssl, Ubuntu:25.10: wolfssl, Ubuntu:26.04:LTS: wolfssl

wolfSSL's AVX2-optimized ML-KEM implementation (mlkem_cmp_avx2) compares only 1536 of the 1568 ciphertext bytes during the Fujisaki-Okamoto re-encryption check in ML-KEM-1024 decapsulation. Ciphertexts that differ from the expected re-encryption solely in bytes 1536-1567 bypass implicit rejection and are accepted as valid, breaking IND-CCA2 security. An attacker able to submit chosen ciphertexts to a decapsulation oracle that uses a static ML-KEM-1024 key, and to observe whether the genuine shared secret or the implicit-rejection secret was produced, can use this as a plaintext-checking oracle to recover the private key. A proof of concept recovered a full ML-KEM-1024 private key with approximately 98% success using roughly 350 chosen ciphertexts. The flaw is a deterministic logic error and does not rely on timing measurements.