Linear Program Reconstruction in Practice
This work demonstrates a practical vulnerability in privacy systems, which is incremental as it applies known algorithms to real-world scenarios.
The authors successfully executed a linear program reconstruction attack on a production statistical queries system using a real dataset, empirically evaluating the effectiveness of two algorithms under various conditions such as dataset sizes and error rates.
We briefly report on a successful linear program reconstruction attack performed on a production statistical queries system and using a real dataset. The attack was deployed in test environment in the course of the Aircloak Challenge bug bounty program and is based on the reconstruction algorithm of Dwork, McSherry, and Talwar. We empirically evaluate the effectiveness of the algorithm and a related algorithm by Dinur and Nissim with various dataset sizes, error rates, and numbers of queries in a Gaussian noise setting.