Yuke Hu

TL;DR: In this paper , Xu et al. proposed three order-preserving desensitization algorithms satisfying a variant of Local Differential Privacy (LDP), called distance-based LDP (dLDP), to improve the accuracy of tree boosting algorithms satisfying differential privacy under vertical FL.

TL;DR: Wang et al. as mentioned in this paper proposed DP-Flames, a differentially private federated KGE with private selection, which offers a better privacy-utility tradeoff by exploiting the entity-binding sparse gradient property of FKGE and comes with a tight privacy accountant by incorporating the state-of-theart private selection technique.

TL;DR: In this paper , the authors introduce new security notions tailored to the specific privacy requirements of dynamic workloads, such as key-value, range-query, and dynamic workload, and present an efficient construction that progressively enables these workloads while provably mitigating systemwide leakage via a suite of algorithms with tunable privacy-efficiency trade-offs.

TL;DR: This paper analyzes the privacy risks of Key-Value caches in Large Language Model inference, demonstrating three attack vectors that can reconstruct sensitive user inputs. A novel defense mechanism, KV-Cloak, effectively mitigates these risks with minimal performance overhead.