Sequential Random Sampling PIR for DNA-Based Data Storage: From a Single Server to Multiple Colluding Servers

Speaker: 王琛 博士后，以色列理工学院计算机科学系             

Inviter: 孔祥粱 副研究员

Title: Sequential Random Sampling PIR for DNA-Based Data Storage: From a Single Server to Multiple Colluding Servers

Language: Chinese 

Time & Venue: 2026.10.30  15:30-16:20  腾讯会议：138-971-336

Abstract: Protecting user privacy is an important challenge in retrieving data from DNA-based storage systems. We consider a model in which data are stored as unordered DNA strands and accessed through uniform random sampling, without deterministic addressing or server-side computation. These constraints prevent the direct application of conventional private information retrieval (PIR) schemes. In this talk, we present a sequential random sampling PIR framework, in which sampled files are transmitted to the user one at a time and the user decides when to stop. We study both a single server and multiple colluding servers, with download cost measured by the expected number of downloaded files and privacy leakage by the mutual information between the desired file and the response of the servers. For both settings, we derive fundamental lower bounds on download cost as a function of privacy leakage and develop explicit retrieval schemes. We characterize the trade-offs achieved by these schemes and establish their optimality in certain parameter regimes.