PARIS: Research Seal+Sparta
Tunable privacy framework for encrypted database and network queries.
Description
Built a privacy-preserving query system combining SEAL at the database layer with SPARTA at the network layer to provide tunable leakage resistance against access, volume, and timing analysis attacks. Implemented ADJ-ORAM-alpha to obscure record access patterns and adjustable padding buckets to hide exact query result sizes while preserving configurable performance tradeoffs. Designed deferred retrieval and dummy traffic injection mechanisms to break traffic correlation and timing inference across encrypted communications. Developed attack simulation modules to benchmark query recovery and database recovery against greedy adversarial baselines, alongside traffic correlation attack testing for network-layer leakage. Integrated parameter sweeping across alpha and padding factors to identify Pareto-optimal security-performance configurations, maintaining total compute overhead below 32x while satisfying all leakage constraints.