Characterizing and Understanding Software Security Vulnerabilities in Machine Learning Libraries

TL;DR

This paper analyzes security vulnerabilities in five major ML libraries, characterizing their types, causes, symptoms, and fixes, and introduces DeepMut, a tool to evaluate test suite adequacy against these vulnerabilities.

Contribution

It provides the first comprehensive characterization of ML library vulnerabilities and develops DeepMut, an automated mutation testing tool based on these insights.

Findings

596 security-related commits analyzed

DeepMut found over 1,000 mutants not covered by existing tests

Insights into vulnerability types, causes, and fixing patterns

Abstract

The application of machine learning (ML) libraries has been tremendously increased in many domains, including autonomous driving systems, medical, and critical industries. Vulnerabilities of such libraries result in irreparable consequences. However, the characteristics of software security vulnerabilities have not been well studied. In this paper, to bridge this gap, we take the first step towards characterizing and understanding the security vulnerabilities of five well-known ML libraries, including Tensorflow, PyTorch, Sickit-learn, Pandas, and Numpy. To do so, in total, we collected 596 security-related commits to exploring five major factors: 1) vulnerability types, 2) root causes, 3) symptoms, 4) fixing patterns, and 5) fixing efforts of security vulnerabilities in ML libraries. The findings of this study can assist developers in having a better understanding of software security vulnerabilities across different ML libraries and gain a better insight into their weaknesses of them. To make our finding actionable, we further developed DeepMut, an automated mutation testing tool, as a proof-of-concept application of our findings. DeepMut is designed to assess the adequacy of existing test suites of ML libraries against security-aware mutation operators extracted from the vulnerabilities studied in this work. We applied DeepMut on the Tensorflow kernel module and found more than 1k alive mutants not considered by the existing test suits. The results demonstrate the usefulness of our findings.