Montage: A Neural Network Language Model-Guided JavaScript Engine Fuzzer

TL;DR

Montage is a novel neural network language model-guided fuzzer that effectively detects JavaScript engine vulnerabilities by transforming ASTs into sequences for training, outperforming previous methods and discovering multiple real-world bugs.

Contribution

This paper introduces Montage, the first NNLM-guided JavaScript fuzzer that transforms ASTs into sequences for effective vulnerability detection.

Findings

Montage found 37 real-world bugs, including three CVEs.

It outperforms previous fuzzing techniques in vulnerability discovery.

Montage can generate valid JavaScript tests from AST sequences.

Abstract

JavaScript (JS) engine vulnerabilities pose significant security threats affecting billions of web browsers. While fuzzing is a prevalent technique for finding such vulnerabilities, there have been few studies that leverage the recent advances in neural network language models (NNLMs). In this paper, we present Montage, the first NNLM-guided fuzzer for finding JS engine vulnerabilities. The key aspect of our technique is to transform a JS abstract syntax tree (AST) into a sequence of AST subtrees that can directly train prevailing NNLMs. We demonstrate that Montage is capable of generating valid JS tests, and show that it outperforms previous studies in terms of finding vulnerabilities. Montage found 37 real-world bugs, including three CVEs, in the latest JS engines, demonstrating its efficacy in finding JS engine bugs.