Skip to content

Latest commit

 

History

History
27 lines (20 loc) · 948 Bytes

File metadata and controls

27 lines (20 loc) · 948 Bytes

IMDb/JOB Benchmark

imdb_benchmark exercises PAC and elastic DP on an IMDb/JOB-style workload with person-level privacy. It is an additional development benchmark, not part of the paper result matrix.

Quick Check

cmake --build build/release --target imdb_benchmark
build/release/extension/privacy/imdb_benchmark --micro --out /tmp/imdb_micro.csv

Full Workload

build/release/extension/privacy/imdb_benchmark \
  --db imdb.db \
  --queries benchmark/imdb/queries.sql \
  --out benchmark/results/imdb.csv

The runner downloads the JOB IMDb data from the CedarDB mirror when needed. Use --data-dir or --url to override the input location.

The privacy unit is name.id: people are sensitive entities, while titles are public objects. Queries in benchmark/imdb/queries_pac_join.sql use broader title joins and may be rejected by elastic DP when the joins do not form one linear privacy-link chain to name.