A study relating to the PETRAS’ Sec-QBS project has been published by authors Shubham Jain, Ana-Maria Crețu, and Yves-Alexandre de Montjoye from Imperial College London. The paper, titled “Adversarial Detection Avoidance Attacks: Evaluating the robustness of perceptual hashing-based client-side scanning”, examines the robustness of client-side scanning in E2EE (End-to-end encryption) communications.
Client-side scanning refers to a piece of software that enables users to monitor illegal activities on their devices. The aim of the publication is to propose a first framework that evaluates the robustness of perceptual hashing-based client-side scanning to detection avoidance attacks. While E2EE enables private communication, its widespread adoption has raised concerns about illegal content being shared undetected. Following the global pushback against key escrow systems, client-side scanning is likely to be a major IoT and privacy topic in the coming year.
The authors of the study tested the robustness of hashing-based client-side scanning and found that more than 99.9% of images could ‘fool’ the algorithms and avoid detection while preserving the content of the image.
A post relating to the study has been published on the Imperial College London’s website.
Read the study here.