Beyond Fidelity: Explaining Vulnerability Localization of Learning-based Detectors

TL;DR

This paper evaluates explanation methods for deep learning-based vulnerability detectors, revealing that current approaches often lack precision in identifying critical vulnerability-related code lines and fidelity metrics are insufficient for assessment.

Contribution

It provides an in-depth evaluation of ten explanation approaches for vulnerability detectors, highlighting their limitations in accurately pinpointing vulnerability-critical code features.

Findings

Fidelity metrics fluctuate significantly across datasets and detectors.

All explanation approaches show poor accuracy in identifying vulnerability-related code lines.

Current explainers are inefficient in selecting important features and are affected by irrelevant artifacts.

Abstract

Vulnerability detectors based on deep learning (DL) models have proven their effectiveness in recent years. However, the shroud of opacity surrounding the decision-making process of these detectors makes it difficult for security analysts to comprehend. To address this, various explanation approaches have been proposed to explain the predictions by highlighting important features, which have been demonstrated effective in other domains such as computer vision and natural language processing. Unfortunately, an in-depth evaluation of vulnerability-critical features, such as fine-grained vulnerability-related code lines, learned and understood by these explanation approaches remains lacking. In this study, we first evaluate the performance of ten explanation approaches for vulnerability detectors based on graph and sequence representations, measured by two quantitative metrics including fidelity and vulnerability line coverage rate. Our results show that fidelity alone is not sufficient for evaluating these approaches, as fidelity incurs significant fluctuations across different datasets and detectors. We subsequently check the precision of the vulnerability-related code lines reported by the explanation approaches, and find poor accuracy in this task among all of them. This can be attributed to the inefficiency of explainers in selecting important features and the presence of irrelevant artifacts learned by DL-based detectors.