On the Design of Privacy-Aware Cameras: a Study on Deep Neural Networks

TL;DR

This paper investigates how camera distortions can be used to prevent neural networks from extracting sensitive personal data, proposing a privacy-aware camera design that balances privacy with data utility.

Contribution

It demonstrates a method to create privacy-aware cameras using image distortions that block sensitive data extraction while preserving non-sensitive information.

Findings

Distorted images prevent license plate recognition.

Non-sensitive data remains extractable from distorted images.

Proposed approach enhances privacy in camera systems.

Abstract

In spite of the legal advances in personal data protection, the issue of private data being misused by unauthorized entities is still of utmost importance. To prevent this, Privacy by Design is often proposed as a solution for data protection. In this paper, the effect of camera distortions is studied using Deep Learning techniques commonly used to extract sensitive data. To do so, we simulate out-of-focus images corresponding to a realistic conventional camera with fixed focal length, aperture, and focus, as well as grayscale images coming from a monochrome camera. We then prove, through an experimental study, that we can build a privacy-aware camera that cannot extract personal information such as license plate numbers. At the same time, we ensure that useful non-sensitive data can still be extracted from distorted images. Code is available at https://github.com/upciti/privacy-by-design-semseg .