Encrypted Data Skipping in DuckDB Iceberg
MSc thesis on a novel Bloom-filter based encrypted data skipping scheme, implemented in DuckDB Iceberg.
My master's thesis supervised by Peter Boncz at CWI (Dutch Center for Mathematics and Computer Science). Thesis covers implemenatation of a novel Bloom-filter based encrypted data skipping scheme in DuckDB Iceberg. This scheme allows range queries (numerical & strings) over encrypted data, without decryption of the metadata. A number of comparisons with current SOTA encrypted querying methods are given. The implementation amkes extensive use of SIMD operations and vectorizations. Microbenchmarks were used extensively to evaluate optimizations.