GENOME RECONSTRUCTION IN BEACONS USING SUMMARY STATISTICS
Kousar Saleem
Master Student
(Supervisor: Asst.Prof.Sinem Sav) Computer Engineering Department
Bilkent University
Abstract: Genomic data-sharing beacons are designed to balance individual privacy with the promotion of scientific discovery; however, they remain critically vulnerable to sophisticated genome reconstruction attacks that exploit publicly released summary statistics. The core vulnerability arises from the beacon protocol’s failure to account for linkage disequilibrium (LD), which enables adversaries to infer individual-level genomic information from aggregated statistics. This work systematically advances the study of genome reconstruction attacks by establishing the feasibility of the attack first using a greedy approach that leverages AF information. Building on this foundation, the research proposes advanced attack frameworks based on optimization formulation that simultaneously capture SNP correlation structure and allele frequency consistency. The proposed approaches improve both the effectiveness and practicality of genome reconstruction attacks.Overall, this work demonstrates that beacon protocols remain fundamentally vulnerable to reconstruction strategies, highlighting the urgent need for robust and adaptive defense mechanisms that explicitly account for genomic dependencies to ensure the long-term privacy of genomic data-sharing infrastructures.
DATE: January 12, Monday @ 11:30 Place: EA 409