Safe crime detection

Abstract

The paper proposes a way to perform surveillance that only invades the privacy of criminals or terrorists, leaving the innocent unsurveilled. This is achieved by combining homomorphic encryption with deep learning to create encrypted neural networks that can be used to detect criminal activity. The paper provides a prototype implementation of such a system using a spam detection dataset, showing how encrypted neural networks can be used to make predictions on encrypted data without revealing the underlying data or the model itself. The paper argues that this approach has the potential to address the trade-off between privacy and security, as it allows for effective crime detection without requiring the collection and analysis of vast amounts of personal data. – AI-generated abstract.