Static Analysis for Regular Expression Denial-of-Service Attacks

TL;DR

This paper introduces a static analysis method to detect regular expressions vulnerable to exponential-time denial-of-service attacks, providing precise vulnerability localization and malicious input generation with efficient performance.

Contribution

It presents a theoretically grounded static analysis tool that outperforms fuzzing in speed and accuracy for identifying vulnerable regexes and generating test inputs.

Findings

Successfully identified vulnerable regexes in large repositories

Analysis runs in seconds, enabling rapid security assessments

Provides malicious input examples for security testing

Abstract

Regular expressions are a concise yet expressive language for expressing patterns. For instance, in networked software, they are used for input validation and intrusion detection. Yet some widely deployed regular expression matchers based on backtracking are themselves vulnerable to denial-of-service attacks, since their runtime can be exponential for certain input strings. This paper presents a static analysis for detecting such vulnerable regular expressions. The running time of the analysis compares favourably with tools based on fuzzing, that is, randomly generating inputs and measuring how long matching them takes. Unlike fuzzers, the analysis pinpoints the source of the vulnerability and generates possible malicious inputs for programmers to use in security testing. Moreover, the analysis has a firm theoretical foundation in abstract machines. Testing the analysis on two large repositories of regular expressions shows that the analysis is able to find significant numbers of vulnerable regular expressions in a matter of seconds.