Differential Privacy in Dynamic Encryption: Balancing Security and Utility in Cloud-Based Systems

With the rapid development of technologies, treating data with confidentiality protection for cloud systems treating data with the guarantee of confidentiality and integrity, so that analytics are meaningful and responsive, is inevitable. The paper argues that Differential Privacy (DP) and Dynamic Encryption (DE) are two promising complementary approaches to secure the cloud system. Specifically, DP codes inference risk from released results while DE mitigates unauthorized access with fine- grained encryption, key rotation, and limited ciphertexts computation. To be more precise, the paper proposes a deployable using both DP and DE reference architecture for the cloud data platform, with a purpose that realizing the optimization of utility–latency–security. Then experimental evaluation based on aggregate Structured Query Language (SQL) queries and Differentially Private Stochastic Gradient Descent (DP-SGD) training will be briefly outlined. Recent work shows that hour-scale (or continuous) key rotation with moderate privacy budgets is able to provide a pragmatic Pareto balance for common query-heavy applications. Specifically, more homomorphic operations can be applied on low depth aggregates to improve security. By merging these two totally different but beneficial fields, it is believed that an insightful view will be explored on the development of secure-by-design and practically meaningful cloud data platforms.