Privacy-Enhancing Computation (PEC) in Surveillance AI

The Core Idea

Privacy-Enhancing Computation (PEC) in Surveillance AI represents a new approach to analyzing surveillance data that ensures the privacy of individuals and entities involved.

PEC Techniques – Secure Multiparty Computation, Homomorphic Encryption, and Trusted Execution Environments

PEC techniques—secure multiparty computation, homomorphic encryption, and trusted execution environments—enable analytics on sensitive data without exposing raw footage or identities. These methods allow multiple parties to perform computations on their private inputs while keeping the data confidential.

Secure multiparty computation allows multiple parties to compute a function over their input data without revealing individual contributions. Homomorphic encryption enables computations to be performed directly on encrypted data, ensuring that only authorized entities can access the results. Trusted execution environments provide secure and isolated computing environments for running applications with strict privacy controls.

Use Cases Include Cross-Agency Analytics and Third-Party Audits Where Data Sharing is Constrained

Use cases include cross-agency analytics and third-party audits where data sharing is constrained. For instance, different agencies can collaborate on a joint analysis without revealing sensitive information about their respective datasets.

Performance tradeoffs require careful engineering and hardware acceleration to ensure that the computational processes remain efficient even when dealing with large volumes of surveillance data.

Frequently asked questions