In plain words
The paper tests whether post-quantum cryptography can make secure TLS 1.3 connections on small devices without taking too much time or energy. The authors tried different combinations of key exchange and identity-checking methods in the wolfSSL library on four devices.
In their tests, ML-KEM-512 handshakes had lower median latency than X25519 handshakes when the identity-checking method was kept the same. The finding applies to the tested library and devices; larger ML-KEM settings did not always have the same speed advantage.
The research
If encrypted information sent by a small connected device is recorded today, could a future quantum computer expose it by breaking classical key exchange?
The paper asks whether post-quantum methods can fit the speed and energy limits of these devices, which often have fewer computing resources than conventional computers.
The authors compare key exchange choices in TLS 1.3 across four devices, finding ML-KEM-512 faster than X25519 in matched-authentication tests.
The authors compared TLS 1.3 handshakes in wolfSSL across four devices, varying key exchange while matching the paired authentication method. They measured latency and energy, with timing limited to the handshake itself.
ML-KEM-512 had lower median latency than X25519 on each tested device in these matched comparisons. The ESP32 result was especially large, and the authors link its unusually slow X25519 performance to missing Xtensa optimization in the tested wolfSSL build.
These results cover the tested wolfSSL configurations and hardware under WiFi conditions. They do not establish production readiness or measure communication costs separately on the four devices.
Analysis
Deep
- Original paper authors
- Olivier Gillot, William J. Buchanan, Callum Turino
- Submitted to arXiv
- License
- arXiv non-exclusive