LSH Then Lock

A Rust implementation of a fuzzy extractor using locality-sensitive hashing (LSH). This tool provides functionality for analyzing and implementing template protection through smart bit selection and correlation analysis.

Warning

⚠️ Research Use Only

This implementation was built specifically to produce data for our research paper. While it's a decent implementation for other researchers to use in their academic work, it is not recommended for production use. The code has not been audited for security vulnerabilities and may not implement all necessary safeguards for real-world applications.

If you plan to use fuzzy extractors in a production environment, you should seek or create a properly audited implementation. If you create a production-ready version of this code and can verify its security, please let me know and I'll be happy to link to your repository.

Installation

Ensure you have Rust and Cargo installed on your system. Then clone this repository:

git clone https://github.com/whyamey/lsh-lock
cd lsh-lock
cargo build --release

Usage

The tool provides several commands for different operations:

Random Sampling

Generate lockers without considering zeta values:

cargo run -- random-sampling -o output.bin -c 250000 -s 80 -d 1024

Options:

• -o, --output: Output file path
• -c, --count: Number of samples (default: 250000)
• -s, --size: Size of each sample (default: 80)
• -d, --dimensions: Feature dimensions (default: 1024)

Zeta Sampling

Generate lockers with respect to zeta values:

cargo run -- zeta-sampling -o output.bin -c confidence.txt -c 250000 -s 80 -a 1.0 --method exponent -d 1024

Options:

• -o, --output: Output file path
• -c, --confidence: Path to confidence file
• -c, --count: Number of samples (default: 250000)
• -s, --size: Size of each sample (default: 80)
• -a, --alpha: Alpha parameter (default: 1.0)
• -d, --dimensions: Feature dimensions (default: 1024)
• --method: Sampling method (gaps/like/ratio/exponent)
• --bad-indices: Comma-separated list of indices to exclude

Correlation Analysis

Generate confidence values by analyzing correlations:

cargo run -- correlate -i input_dir -o output.txt -n 100 -m both -d 1024

Options:

• -i, --input: Input directory path
• -o, --output: Output file path
• -n, --num-files: Number of files to analyze (default: 100)
• -m, --mode: Analysis mode (single/pairs/both)
• -d, --dimensions: Feature dimensions (default: 1024)

Entropy Analysis

Analyze the entropy of generated lockers:

cargo run -- analyze -i indices.bin -t templates_dir -n 1000

Options:

• -i, --input: Input file containing indices
• -t, --templates: Templates directory path
• -n, --count: Number of samples to analyze (default: 1000)

Cosine LSH Generation

Generate random projection LSH lockers using cosine distance:

cargo run -- cosine-generate -o output.bin -c 250000 -s 60

Options:

• -o, --output: Output file path
• -c, --count: Number of lockers to generate (default: 250000)
• -s, --size: Size of each locker (default: 60)

Cosine LSH Analysis

Analyze entropy using cosine LSH:

cargo run -- analyze-cosine -i input.bin -t templates_dir -n 1000

Options:

• -i, --input: Input file containing cosine lockers
• -t, --templates: Templates directory path
• -n, --count: Number of samples to analyze (default: 1000)

TAR Analysis

Calculate True Accept Rate for the lockers:

cargo run -- tar -i indices.bin -t templates_dir -n 250000

Options:

• -i, --input: Input file containing indices
• -t, --templates: Templates directory path
• -n, --count: Number of samples to analyze (default: 250000)

Cosine TAR Analysis

Calculate True Accept Rate for cosine LSH lockers:

cargo run -- tar-cosine -i input.bin -t templates_dir -n 250000

Options:

• -i, --input: Input file containing cosine lockers
• -t, --templates: Templates directory path
• -n, --count: Number of samples to analyze (default: 250000)

Multi-Template TAR Analysis

Find TAR with multiple template matching attempts:

cargo run -- tar-multi -i indices.bin -t templates_dir -n 250000 -t 10 -b 1 --input-selection strategy --output-selection strategy

Options:

• -i, --input: Input file containing indices
• -t, --templates: Templates directory path
• -n, --count: Number of samples to analyze (default: 250000)
• -t, --tries: Number of template attempts (default: 10)
• -b, --base: Number of base templates (default: 1)
• --input-selection: Strategy for input template selection
• --output-selection: Strategy for output template selection

Citation

If you use our work, please cite:

@misc{cryptoeprint:2024/100,
      author = {Sohaib Ahmad and Sixia Chen and Luke Demarest and Benjamin Fuller and Caleb Manicke and Alexander Russell and Amey Shukla},
      title = {Fuzzy Extractors are Practical: Cryptographic Strength Key Derivation from the Iris},
      howpublished = {Cryptology {ePrint} Archive, Paper 2024/100},
      year = {2024},
      url = {https://eprint.iacr.org/2024/100}
}