Software Vulnerability Detection Using a Lightweight Graph Neural Network

Miles Farmer, Ekincan Ufuktepe, , Anne Watson, Hialo Muniz Carvalho, Vadim Okun, , Zineb Maasaoui, Kannappan Palaniappan · 2026

Read Software Vulnerability Detection Using a Lightweight Graph Neural Network by Miles Farmer, Ekincan Ufuktepe, , Anne Watson, Hialo Muniz Carvalho, Vadim Okun, , Zineb Maasaoui, Kannappan Palaniappan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-29216.txt.

Description

Farmer and co-authors present VulGNN, a lightweight graph neural network for finding software flaws. Language models perform well at this task but need heavy compute. Code has a natural graph structure, which VulGNN exploits. The model nearly matches language-model performance. Yet it is about 100 times smaller and quick to retrain. The authors describe the architecture and run ablation studies. They also test how well it generalizes across code datasets. VulGNN suits edge deployment in real development pipelines.

Catalog record

CREATORMiles Farmer, Ekincan Ufuktepe, , Anne Watson, Hialo Muniz Carvalho, Vadim Okun, , Zineb Maasaoui, Kannappan Palaniappan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46741
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.29216

Browse the full library · About the Tower of Babel