Beyond TVLA: Anderson-Darling Leakage Assessment for Neural Network Side-Channel Leakage Detection

2026-03-19

AI Summary

This work addresses the limitations of traditional Test Vector Leakage Assessment (TVLA), which relies solely on mean differences and struggles to detect higher-order distributional leakage in the presence of countermeasures such as masking and random shuffling. To overcome this, the paper introduces the Anderson-Darling Leakage Assessment (ADLA) framework, which, for the first time, incorporates the two-sample Anderson-Darling test into side-channel analysis. By comparing full cumulative distribution functions rather than just means, ADLA transcends TVLA’s inherent constraints. Experimental validation on the ChipWhisperer-Husky platform, integrating multilayer perceptrons with standard protection mechanisms, demonstrates that ADLA significantly enhances sensitivity to subtle or high-order leakage in protected neural network implementations—even with a limited number of power traces—thereby confirming its practical superiority in real-world security evaluations.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Measurement, analysis, and circumvention of Web censorship

Abstract

Test Vector Leakage Assessment (TVLA) based on Welch's $t$-test has become a standard tool for detecting side-channel leakage. However, its mean-based nature can limit sensitivity when leakage manifests primarily through higher-order distributional differences. As our experiments show, this property becomes especially crucial when it comes to evaluating neural network implementations. In this work, we propose Anderson--Darling Leakage Assessment (ADLA), a leakage detection framework that applies the two-sample Anderson--Darling test for leakage detection. Unlike TVLA, ADLA tests equality of the full cumulative distribution functions and does not rely on a purely mean-shift model. We evaluate ADLA on a multilayer perceptron (MLP) trained on MNIST and implemented on a ChipWhisperer-Husky evaluation platform. We consider protected implementations employing shuffling and random jitter countermeasures. Our results show that ADLA can provide improved leakage-detection sensitivity in protected implementations for a low number of traces compared to TVLA.

Problem

• side-channel leakage
• neural network
• TVLA
• distributional differences
• leakage detection

Innovation

• Anderson-Darling test
• side-channel leakage detection
• neural network security
• TVLA
• distributional difference