"New AI Model Can Help Prevent Damaging and Costly Data Breaches"

Privacy experts at Imperial College London have developed an Artificial Intelligence (AI) algorithm capable of automatically testing privacy-preserving systems for potential data leaks. According to the researchers, this is the first time AI has been used to automatically find vulnerabilities in this type of system, which is used by Google Maps, Facebook, and more. Imperial's Computational Privacy Group analyzed attacks on Query-Based Systems (QBS). These systems are controlled interfaces through which analysts can query data to extract useful aggregate information about the world. Then they created QuerySnout, a new AI-enabled method, to detect attacks on QBS. QBS provides analysts access to statistical collections derived from individual-level data such as location and demographics. They are currently used in Google Maps to display real-time information on how busy an area is, as well as in Facebook's Audience Measurement feature to estimate audience size in a specific location or demographic to improve advertising promotions. The team discovered that powerful and accurate attacks against QBS can be easily detected with the press of a button. QuerySnout learns which questions to ask the system to obtain answers. It then learns to automatically combine the responses in order to detect potential privacy vulnerabilities. The model can create an attack using Machine Learning (ML). The attack involves a collection of queries that combines the answers to reveal specific private information. This process is completely automated and uses the 'evolutionary search' method to enable the QuerySnout model to discover the appropriate questions to ask. This process takes place in a 'black-box setting,' meaning the AI only needs access to the system but does not need to understand how it works to detect vulnerabilities. This article continues to discuss the new AI algorithm developed by privacy experts at Imperial London College to automatically test privacy-preserving systems for potential data leaks. 

Imperial College London reports "New AI Model Can Help Prevent Damaging and Costly Data Breaches"