HetCSS Experiment Data: HNSW-ORAM Benchmark Datasets (Individual Files)

This dataset contains pre-built HNSW index files and query vectors for reproducing end-to-end experiments in the HetCSS project (Heterogeneous Confidential Similarity Search with ORAM protection). Datasets included: - sift1m: 1M vectors, M=64, dim=128 - laion100K: 100K vectors, M=64, dim=512 - tripclick: 1.5M vectors, M=128, dim=768 - msmarco_bert: 8.8M vectors, M=128, dim=768 File naming convention: - {dataset}_query.fvecs: Query vectors - {dataset}_index.bin: Pre-built HNSW index To use with HetCSS: 1. Create data/{dataset}/ folders 2. Rename files: {dataset}_query.fvecs -> query.fvecs, {dataset}_index.bin -> index.bin Or use the download_data.sh script from the HetCSS repository which handles this automatically.