A Contributory Public-Event Recording and Querying System

Abstract

CCTV (Closed-Circuit Television) systems are commonly used for security and surveillance. They provide a visual record of events, which can be used to monitor criminal activity, support investigations, and improve public safety. Many cities have implemented a number of cameras for surveillance, with Delhi, India having 1446 cameras per square mile. These cameras are installed mainly by official traffic police or city authorities, but private organizations and individuals also have their cameras pointed toward public spaces. In many cases, the city authorities or police require query capability and access of the video feed from these CCTV cameras to perform diligent inquiries but the private entities usually do not share the raw footage due to various concerns. To address the above issue, we introduce the “Public Event Recording and Querying System (PERQS)”, a video analysis-based querying system. PERQS is a contributory CCTV network that encompasses both private and public parties. PERQS provides a reliable and secure solution for querying CCTV video feeds by leveraging video analytic algorithms. It ensures video privacy by allowing participants to perform video analysis locally on their servers. Additionally, PERQS employs hash-based commitments and query consensus, guaranteeing tamper-proof and accurate results. Furthermore, it provides a novel consensus mechanism called time-travel consensus to build trust in the query output even in the case of a lack of cameras pointing at a particular location. We introduce a query language PERQL, similar to SQL, which makes the system expandable and flexible. The plug-and-play architecture allows for integration with advanced vision models and algorithms for analysis.