Privacy-Preserving Video Classification with Convolutional Neural Networks

TL;DR

This paper presents a novel privacy-preserving video classification method using convolutional neural networks that ensures user data and model confidentiality through secure multi-party computation, achieving high accuracy in emotion recognition.

Contribution

It introduces new MPC protocols for oblivious frame selection and label aggregation, enabling end-to-end privacy-preserving video classification without revealing sensitive data or model parameters.

Findings

Achieves state-of-the-art accuracy in private emotion recognition

Works securely under various adversarial settings

Maintains privacy of user videos and model parameters

Abstract

Many video classification applications require access to personal data, thereby posing an invasive security risk to the users' privacy. We propose a privacy-preserving implementation of single-frame method based video classification with convolutional neural networks that allows a party to infer a label from a video without necessitating the video owner to disclose their video to other entities in an unencrypted manner. Similarly, our approach removes the requirement of the classifier owner from revealing their model parameters to outside entities in plaintext. To this end, we combine existing Secure Multi-Party Computation (MPC) protocols for private image classification with our novel MPC protocols for oblivious single-frame selection and secure label aggregation across frames. The result is an end-to-end privacy-preserving video classification pipeline. We evaluate our proposed solution in an application for private human emotion recognition. Our results across a variety of security settings, spanning honest and dishonest majority configurations of the computing parties, and for both passive and active adversaries, demonstrate that videos can be classified with state-of-the-art accuracy, and without leaking sensitive user information.