FIDO SPHINCS+ Experiments

This is the meta repository for the code accompanying the paper Stateless Hash-Based Signatures for Post-Quantum Security Keys. It ties together all scripts and code needed to reproduce the results.

Structure

The relevant code is distributed amongst four repositories:

• sphincsplus - The SPHINCS+ C code, with a tool instantiate and test different parameter sets.
• sphincs_wrapper - The Rust wrapper for SPHINCS+
• OpenSK - OpenSK modified to use Sphincs
• FIDO SPHINCS+ Experiments - This meta repository with benchmarking scripts

The dependencies are included as submodules.

Reproduce

1. Repository and Software

Clone the repo with all submodules:

git clone --recurse-submodules https://github.com/rugo/fido-sphincs-experiments

You need the JLink software for the speed and stack benchmarks. A recent Python version is needed for the helper scripts.

2. Hardware

To reproduce the results of the paper, you need an nRF52840 development kit. Connect both USB cables to a host computer.

3. Parameter file

The parameter file needs to have one SPHINCS+ parameter set per line. Each line needs to contain the following, space separated values:

1. Signature size: Given in bytes and rows are sorted by this ascending.
2. Signing speed: Number of hash calls for signing. The search is limited to < 10^9 calls, which corresponds to roughly < 1 min signing on a modern CPU.
3. Verification speed: Number of hash calls for verification.
4. Probability for FORS forgery after 2^y signatures.
5. The parameter h in SPHINCS+.
6. The parameter d in SPHINCS+.
7. The parameter b in SPHINCS+.
8. The parameter k in SPHINCS+.
9. The parameter w in SPHINCS+.
10. Security degradation: The value z such that after 2^z signatures, the forgery probability is still < 2^{-112, -128, -192} (for a security target of 128, 192, 256 bits)

Although (3) and (10) are not used by the tooling. This format is compatible to the one used here. `

4. Benchmarking

Run the run_benchmark.py script with a parameter file of your choice.

Like so:

python3 run_benchmarks.py params/128_10.txt 10-sig

The results will be stored in the results/10-sig folder.

Unfortunately you have to press the board's reset button before the FIDO HID benchmark (make/get) starts. Do so once the script says "Configuring device". This is not needed for the stack and speed benchmarks.

5. Pretty Printing

To export the results into one file, use the print_results.py script on the result folder:

p3 print_results.py results/10-sig

This will output a csv file.