Breaking the Black Box: Byte-Level Boundary Inference of Real-World Antivirus Systems

2026-09-25

AI Summary

This study addresses the limitation of existing antivirus (AV) software, which typically yields only binary verdicts without providing fine-grained explanations of the critical byte regions triggering detection. To bridge this gap, this work proposes AVHunter, a black-box framework that systematically probes eleven real-world AV engines to construct a large-scale dataset and trains specialized deep learning models to localize key byte boundaries and reproduce detection logic. As the first approach to achieve byte-level critical region inference for real-world AV systems, this research reveals a novel risk of AV knowledge leakage. Experimental results demonstrate that the trained models attain an average boundary prediction recall of 85.07% and a 97.43% consistency rate with target AV detections, while the inferred detection logic remains stable over a seven-month period.

Abstract

Existing approaches for understanding the detection logic of real-world antivirus (AV) software infer only binary malware/benign decisions from black-box queries, providing limited insight into the fine-grained decision-critical regions that govern AV detection. In this paper, we present \textbf{AVHunter}, the first framework for inferring byte-level decision-critical regions of real-world AV products under a black-box threat model. AVHunter constructs the first large-scale Byte-Level AV Boundary Dataset (BABD) by systematically probing 11 real-world AV products, revealing that modern AV detections are largely associated with a small number of compact decision-critical byte regions. Leveraging BABD, AVHunter trains AV-specific models that not only reproduce binary AV decisions, but also localize the decision-critical byte regions underlying these decisions, achieving an average boundary prediction recall of 85.07% while maintaining 97.43% detection agreement with the target AVs. We further validate that the predicted regions capture genuine AV decision knowledge through boundary-guided malware evasion, false-positive induction on benign executables, and a seven-month longitudinal study demonstrating that the inferred regions remain largely stable as AV products evolve. Overall, AVHunter moves beyond conventional binary-label AV modeling by enabling fine-grained boundary-region localization and revealing a new form of AV knowledge leakage with important implications for malware analysis, AV security, and boundary-aware defenses.

Problem

• Antivirus Detection Logic
• Black-box Inference
• Byte-level Boundary
• Decision-critical Regions
• Malware Analysis

Innovation

• Byte-Level Boundary Inference
• Black-Box Antivirus Analysis
• Decision-Critical Regions
• Boundary Dataset
• Malware Evasion