.png){kind=icon}
.png){kind=icon}
Return
Detecting WebAssembly Runtime Bugs With Grammar-Guided Program Mutation
.png){kind=icon}
PRE
AI
DOI：10.1109/TR.2025.3614352.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
WebAssembly has been steadily gaining popularity in recent years, as an industry-wide collaborative effort to combine near-native performance with the convenience of high-level languages. It is crucial to establish effective testing methods for WebAssembly runtime, since all WebAssembly programs depend on the quality of the runtime. Meanwhile, fuzz testing is an effective approach for detecting bugs in compilers and runtimes, and has achieved promising results. However, we still face two major challenges in applying fuzz testing to detect bugs in WebAssembly runtime, i.e., the valid test program challenge and the diverse test program challenge. We propose Wagen, a grammar-guided program mutation approach. To tackle the valid test program challenge, Wagen incorporates a set of custom mutation rules designed to mutate the leaf nodes and subtrees of these syntax trees to iteratively generate new WebAssembly program variants. To tackle the diverse test program challenge, Wagen utilizes coverage information, and employs the Markov Chain Monte Carlo sampling to guide the mutation rule selection procedure. Finally, Wagen uses differential testing to identify WebAssembly runtime bugs. Extensive experiments demonstrate that Wagen outperforms the existing techniques in terms of bug detection capability and code coverage. Within six-month evaluations, we have reported 34 bugs on five widely used open-source WebAssembly runtimes, of which 18 have been confirmed.
Keywords:
Webassembly
Runtime
Computer bugs
Testing
Fuzzing
Codes
Syntactics
Security
Browsers
Software
Differential testing
fuzz testing
grammar-guided mutation
test program generation
WebAssembly runtime testing
Journal
IF:
5.7
Papers:
2.7K
Citations:
8.5K