SAGE: Signal-Amplified Guided Embeddings for Vulnerability Detection
Open Access
- 1 October 2026
- journal article
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Software Engineering
- Vol. 3 (ISSTA) , 1147-1170
Abstract
Software vulnerabilities are a primary threat to modern infrastructure. While static analysis and Graph Neural Networks have long served as the foundation for vulnerability detection, the emergence of Large Language Models (LLMs) has introduced a transformative paradigm driven by superior semantic reasoning and cross-environment generalization. However, in the context of LLM-based vulnerability detection, we identify a fundamental bottleneck in these models termed Signal Submersion : a state where features related to vulnerability are activated internally but numerically overwhelmed by dominant functional semantics. To address this, we propose SAGE ( S ignal- A mplified G uided E mbeddings), a framework that shifts from passive signal submersion to active signal recovery. SAGE integrates task-conditional Sparse Autoencoders (SAEs) to isolate and amplify these faint vulnerability signals. Extensive evaluations on BigVul, PrimeVul, and PreciseBugs demonstrate that SAGE achieves state-of-the-art performance. Notably, SAGE mitigates Signal Submersion by increasing the internal Signal-to-Noise Ratio (SNR) by 12.7× via sparse manifold projection. This mechanistic intervention enables a 7B model to achieve up to 318% Matthews Correlation Coefficient (MCC) gains on unseen distributions and a 319% gain on classic datasets. By maintaining robust performance across 13 programming languages and outperforming 34B baselines, SAGE establishes a more efficient and scalable path to software security than simple parameter scaling.Keywords
All Related Versions
Funding Information
- National Natural Science Foundation of China (U25A20425)
- National Natural Science Foundation of China (62232010)
- National Natural Science Foundation of China (62302266)
- National Natural Science Foundation of China (U23A20302)
- National Natural Science Foundation of China (U24A20244)
- National Natural Science Foundation of China (U24B20149)
- Taishan Scholars Program (tsqn202507069)
- Research Project of Quancheng Laboratory (QCL20250106)
- Natural Science Foundation of Shandong Province (ZR2024QF093)
- Young Talent of Lifting engineering for Science and Technology in Shandong (SDAST2025QTB031)
This publication has 52 references indexed in Scilit:
- Vulnerability Detection with Code Language Models: How Far are We?2025
- Steering Large Language Models for Vulnerability Detection2025
- ProRLearn: boosting prompt tuning-based vulnerability detection by reinforcement learningAutomated Software Engineering, 2024
- PreciseBugCollector: Extensible, Executable and Precise Bug-Fix Collection: Solution for Challenge 8: Automating Precise Data Collection for Code Snippets with Bugs, Fixes, Locations, and Types2023
- When Less is Enough: Positive and Unlabeled Learning Model for Vulnerability Detection2023
- Data Quality for Software Vulnerability Datasets2023
- An empirical study of vulnerability discovery methods over the past ten yearsComputers & Security, 2022
- The Balanced Accuracy and Its Posterior Distribution2010
- ITS4: a static vulnerability scanner for C and C++ code2002
- Comparison of the predicted and observed secondary structure of T4 phage lysozymeBiochimica et Biophysica Acta (BBA) - Protein Structure, 1975