Skip to main navigation Skip to search Skip to main content

Optimization Framework for Malware Detection Based on Adversarial Networks and Gradient Reversal

  • Yanchen Qiao
  • , Bowen Li
  • , Weizhe Zhang*
  • , Yu Zhang
  • , Shudong Li
  • *Corresponding author for this work
  • Pengcheng Laboratory
  • Harbin Institute of Technology
  • Guangzhou University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Packing is a prevalent method employed by malware to evade antivirus software. Early on, there were fewer instances of benign software being packed, leading to biases in datasets collected by antivirus vendors. These biases have caused malware detection engines based on deep learning to overly rely on features from the packed regions. With the increasing use of packing techniques for purposes such as copyright protection, more benign software applications are susceptible to being misclassified as malware. Balancing effective malware detection while mitigating false positives for benign packed software is a pressing issue. To address this challenge, we propose an optimization framework for malware detection based on adversarial networks and gradient reversal. By integrating adversarial networks and gradient reversal techniques, this framework diminishes the reliance on packing-related features during malware detection, prompting deep learning models to prioritize features unrelated to packing. Experimental results demonstrate that our framework significantly enhances malware detection accuracy, particularly showcasing robust capabilities in detecting newly packed malware.

Original languageEnglish
Title of host publicationCyberspace Simulation and Evaluation - 3rd International Conference, CSE 2024, Proceedings
EditorsGuangxia Xu, Guangxia Xu, Wanlei Zhou, Jiawei Zhang, Yanchun Zhang, Yan Jia
PublisherSpringer Science and Business Media Deutschland GmbH
Pages295-309
Number of pages15
ISBN (Print)9789819645022
DOIs
StatePublished - 2025
Event3rd International Conference on Cyberspace Simulation and Evaluation, CSE 2024 - Shenzhen, China
Duration: 26 Nov 202428 Nov 2024

Publication series

NameCommunications in Computer and Information Science
Volume2420 CCIS
ISSN (Print)1865-0929
ISSN (Electronic)1865-0937

Conference

Conference3rd International Conference on Cyberspace Simulation and Evaluation, CSE 2024
Country/TerritoryChina
CityShenzhen
Period26/11/2428/11/24

Keywords

  • Adversarial network
  • Gradient reversal
  • Malware
  • Optimization framework

Fingerprint

Dive into the research topics of 'Optimization Framework for Malware Detection Based on Adversarial Networks and Gradient Reversal'. Together they form a unique fingerprint.

Cite this