Fig. 2
From: BLOOM: BLoom filter based oblivious outsourced matchings
General overview of our approach: The data owner holds a database with genomes of Patients P i=1…n which she encodes row-wise as Bloom filters (Step 1) and encrypts and uploads them to the cloud (Step 2) in the preprocessing phase. In the online phase, the data owner encodes, encrypts (only in FHE-BLOOM), and uploads her query. In the encrypted domain, the cloud matches the query to each database record (Step 3) and aggregates the results (Step 4) without ever learning the data in clear by utilizing the homomorphic properties of the chosen encryption scheme. The final results are returned to the data owner, who decrypts (Step 5) and postprocesses the results (Step 6) to obtain a list of patients that match her query