Privacy-Preserving Multimedia Mobile Cloud Computing Using Protective Perturbation

TL;DR

The paper addresses privacy in mobile cloud-based multimedia processing by introducing PMC$^2$, a framework that moves perturbation generation to a secure edge server and uses a neural compressor tailored to perturbed images to dramatically reduce cross-domain bandwidth. The system leverages confidential computing (AMD SEV/CoCo) to protect perturbation generation and a perturbation-aware neural encoder/decoder pair to preserve service accuracy while saving data transmission costs. Experimental results on CIFAR-10 with eight target models show substantial improvements in latency and mobile power, plus bandwidth reductions exceeding 99% compared with traditional PNG/JPEG baselines, while maintaining high recognition accuracy. Overall, PMC$^2$ offers a practical, privacy-preserving, low-bandwidth mobile cloud solution and is slated for open-source release to spur further research.

Abstract

Mobile cloud computing has been adopted in many multimedia applications, where the resource-constrained mobile device sends multimedia data (e.g., images) to remote cloud servers to request computation-intensive multimedia services (e.g., image recognition). While significantly improving the performance of the mobile applications, the cloud-based mechanism often causes privacy concerns as the multimedia data and services are offloaded from the trusted user device to untrusted cloud servers. Several recent studies have proposed perturbation-based privacy preserving mechanisms, which obfuscate the offloaded multimedia data to eliminate privacy exposures without affecting the functionality of the remote multimedia services. However, the existing privacy protection approaches require the deployment of computation-intensive perturbation generation on the resource-constrained mobile devices. Also, the obfuscated images are typically not compliant with the standard image compression algorithms and suffer from significant bandwidth consumption. In this paper, we develop a novel privacy-preserving multimedia mobile cloud computing framework, namely $PMC^2$, to address the resource and bandwidth challenges. $PMC^2$ employs secure confidential computing in the cloud to deploy the perturbation generator, which addresses the resource challenge while maintaining the privacy. Furthermore, we develop a neural compressor specifically trained to compress the perturbed images in order to address the bandwidth challenge. We implement $PMC^2$ in an end-to-end mobile cloud computing system, based on which our evaluations demonstrate superior latency, power efficiency, and bandwidth consumption achieved by $PMC^2$ while maintaining high accuracy in the target multimedia service.

Figures (5)

• Figure 1: On-device power traces with & without privacy-preserving perturbation generation in the mobile cloud image recognition system.
• Figure 2: End-to-end workflow of the proposed PMC$^2$ system.
• Figure 3: Workflow of the perturbation-aware neural compression and decompression.
• Figure 4: Timing performance of the PMC$^2$ system and the baseline protective perturbation system, employing 8 target models. Each target model is deployed in four different ways, including PNG, JPG (JPEG), NC (Neural Compressor), and PP (baseline protective perturbation system using PNG).
• Figure 5: The power performance comparison between the baseline protective perturbation (PP) system and our PMC$^2$ system using different target models. PNG is used in the baseline system, and neural compressor is used in PMC$^2$.