Protecting Patient Privacy in Healthcare Analytics with Fully Homomorphic Encryption and Differential Privacy

Sammanfattning

Data analytics in the healthcare domain requires access to sensitive patient information,creating conflicting interests between the need for usability and privacyrequirements. Fully Homomorphic Encryption (FHE) enables computation on encrypteddata, while Differential Privacy (DP) protects individuals against inferenceattacks by introducing controlled noise into aggregate query results.This thesis investigates the combined use of FHE and DP for privacy-preservinghealthcare analytics and evaluates the resulting privacy guarantees, performance,and practical limitations. Three aggregation queries are implemented and evaluatedin a multi-party privacy-preserving system using multiple FHE schemes andlibraries, including the BFV, BGV, CKKS, and TFHE schemes using the MicrosoftSEAL and Concrete FHE libraries. Performance, accuracy, ciphertext expansion,and compliance to confidentiality and availability requirements are assessed using asynthetic healthcare dataset.The results show that combining FHE and DP strengthens protection against eavesdroppingand membership inference attacks compared to using either of the methodsalone. However, the increased privacy comes at a large cost in performance and usability.Encrypted query execution is orders of magnitude slower than plaintextexecution, and current FHE libraries provide limited support for common statisticaloperations. Additionally, even in the best case, ciphertexts span several megabytesfor a single value, although this can be partially mitigated through compression priorto storage or network transmission. Finally, executing homomorphic computationsin an insecure environment could expose encrypted data to side-channel attacks suchas power measurement or timing attacks.These limitations represent a significant hindrance for large-scale deployment ofFHE in statistical contexts. Improvements to the FHE ecosystem in regards toperformance and usability could enable future large-scale deployment.