PQC Migration Benchmark

PQC Migration Benchmark: An Insightful Analysis of Key Exchange Protocols

The PQC Migration Benchmark project provides a comprehensive performance analysis of key exchange protocols, specifically comparing classical ECDH (X25519), post-quantum ML-KEM (FIPS 203 / Kyber), hybrid X25519+ML-KEM, and ML-DSA-signed ML-KEM key exchange methods. This benchmarking framework operates a real handshake over TCP sockets, leveraging liboqs alongside Python's cryptography library.

Importance of the Project

As the landscape of cryptography evolves, particularly with NIST recently finalizing post-quantum cryptography standards, understanding how these protocols perform is critical. The threat posed by potential quantum computers is not merely theoretical; adversaries could harvest encrypted data today and decrypt it in the future. This reality underscores the necessity for a seamless transition to hybrid key exchange methods that merge classical and post-quantum techniques. Most organizations are currently adopting this approach to enhance security as it requires both key primitives to be compromised to lose confidentiality.

Objectives and Scope of Measurement

The project aims to measure the costs associated with migrating from classical key exchange methods to hybrid and post-quantum approaches, focusing on handshake latency and the amount of data transmitted. The benchmarking methodology is distilled to specific modes, bypassing the complexity of a full TLS 1.3 implementation.

Here are the four modes measured:

Mode	Protects	Server → Client	Client → Server
classical	Confidentiality only	X25519 pubkey	X25519 pubkey
pqc	Confidentiality only	ML-KEM pubkey	ML-KEM ciphertext
hybrid	Confidentiality only	X25519 pubkey ‖ ML-KEM pubkey	X25519 pubkey ‖ ML-KEM ciphertext
signed	Confidentiality and authenticity	ML-DSA pubkey, then ML-KEM pubkey ‖ ML-DSA signature	ML-KEM ciphertext

Each handshake undergoes several assessments:

1. Establishes a real TCP server thread.
2. Connects a client socket and times the entire handshake process.
3. Validates that both parties derive the same session key, ensuring correct operations.
4. Records the handshake time and total bytes exchanged, while incorporating configurable latency to distinguish between compute and network costs.

Key Findings

The project outlines significant findings on the performance impacts of different key exchange protocols:

• Compute Cost Comparison: ML-KEM exhibits comparable performance to classical X25519 with minor overhead differences despite being a different primitive. The overhead increases notably when signing is involved.
• Data Transmission: The wire size scales significantly, with classical methods being a fraction of the size compared to hybrid and signed modes, where data can exceed generational levels multiple times over.
• Latency Contributions: Interestingly, the signed mode introduces an additional round-trip delay, complicating migration timelines for real-world applications, especially with increased network latency scenarios.

Conclusion and Future Directions

For organizations considering the transition to post-quantum cryptography, what this project unveils is critical: while hybrid solutions mitigate risk effectively, the associated costs—both in terms of bandwidth and latency—need consideration in deployment planning. Further exploration is planned in developing a complete TLS 1.3 handshake to better align these benchmarks with real-world cryptographic applications.

Accessing the Results

Detailed results and graphs from all benchmarking runs can be found in the results/ directory, which contains insights from comprehensive trials conducted across various configurations.