Seabed is a project to provide analytics over encrypted Big Data. The challenge is to develop fast yet secure cryptographic techniques that support a suite of applications such as Business Intelligence tools and large-scale Machine Learning frameworks. Towards this, we have developed two novel cryptographic techniques:a high-performance, additive symmetric homomorphic scheme (ASHE) and Splayed ASHE (SPLASHE), an encryption scheme that thwarts frequency analysis-based attacks.
We have built Seabed into Apache Spark and we show that we can perform analytics over encrypted data with an average overhead of only 25%, where previous techniques had overheads ranging between 10x and 100x. Going forward, we are looking at building Machine Learning models over encrypted data.