A Privacy-Preserving Machine Learning Framework for Edge Intelligence: An Empirical Analysis

2026-05-07

AI Summary

This work addresses the critical challenge of balancing data privacy and system efficiency in edge intelligence applications such as smart healthcare and manufacturing. The authors propose a four-layer privacy-preserving machine learning framework and, for the first time, systematically quantify the trade-offs among accuracy, latency, and energy consumption of Differential Privacy (DP), Secure Multi-Party Computation (SMC), and Fully Homomorphic Encryption (FHE) in edge inference through real-world deployment and large-scale trace-based simulation. Results show that DP incurs manageable accuracy degradation under low latency (less than 18% for LeNet and approximately 35% for AlexNet), SMC latency is highly sensitive to bandwidth (30% reduction at 500 Mbps versus 250 Mbps), and FHE introduces roughly a 1000× latency overhead. Additionally, DP significantly reduces the data efficiency of black-box model stealing attacks.

Abstract

As Edge Intelligence (EI) becomes increasingly prevalent in domains such as smart healthcare, manufacturing, and critical infrastructure, ensuring data privacy while maintaining system efficiency is a growing challenge. This paper presents a new privacy-preserving machine learning (PPML) framework tailored for EI applications, including a four-layer system architecture and training and inference algorithms. We focus on three leading approaches: Differential Privacy (DP), Secure Multi-party Computation (SMC), and Fully Homomorphic Encryption (FHE), and assess their impact on key performance metrics, including model accuracy, response time, and energy consumption. Results from real implementation and extensive trace-based simulations of inference tasks show that DP generally preserves throughput and latency close to plaintext baselines, while accuracy drops with model complexity (up to 35 percent on AlexNet and under 18 percent on LeNet for FordA). SMC performance is driven by communication; network bandwidth and round complexity determine end-to-end latency. For AlexNet, increasing link capacity from 250 Mbps to 500 Mbps reduces latency by about 30 percent. FHE is highly sensitive to model structure and numerical precision bit width, with tighter parameters imposing substantial compute overhead; we observe roughly a 1000 times increase in response time compared to DP. Beyond efficiency, DP shifts the privacy-utility-extractability frontier by reducing the attacker's data efficiency in black-box model stealing, whereas SMC and FHE, while protecting inputs and parameters during inference, require complementary output controls to achieve similar resistance to extraction. These findings provide critical insights into the trade-offs between privacy, performance, and resource efficiency in edge computing scenarios.

Problem

• Edge Intelligence
• Privacy-Preserving Machine Learning
• Data Privacy
• System Efficiency
• Performance Trade-offs

Innovation

• Privacy-Preserving Machine Learning
• Edge Intelligence
• Differential Privacy
• Secure Multi-party Computation
• Fully Homomorphic Encryption