X-Wing (Barbosa et al., CiC Volume 1, Issue 1) is a hybrid key encapsulation mechanism (KEM) currently considered for standardization by IETF and deployed by major companies such as Google to ensure a secure transition to post-quantum cryptography. It combines a classical X25519 KEM with the post-quantum ML-KEM-768.
In this paper, we propose the first analysis of the anonymity of X-Wing. We are interested in tight and memory-tight reductions that offer stronger security guarantees. We first establish in the standard model that for any IND-CCA secure KEM, weak anonymity implies full anonymity, and our reduction is tight not only in success probability and time but also in memory consumption. We then prove in the random oracle model that X-Wing achieves weak anonymity if both X25519 and ML-KEM-768 are weakly anonymous. The former can even be proven without a hardness assumption.
Our proof on the weak anonymity of X-Wing does not preserve the memory-tightness of the underlying KEMs. To improve it, we propose a slight variant of X-Wing that preserves memory-tightness. Finally, we improve the existing IND-CCA proof of the original X-Wing by Barbosa et al. using our new memory-tight analysis.