mlsanityData to 7 Oct 2026

MERLIN -- Malware Evasion with Reinforcement LearnINg

arXiv 2203.12980 ↗First posted 24 Mar 2022cs.CR cs.AIcited by 10

Not on the map. Why?

Abstract (arXiv)

In addition to signature-based and heuristics-based detection techniques, machine learning (ML) is widely used to generalize to new, never-before-seen malicious software (malware). However, it has been demonstrated that ML models can be fooled by tricking the classifier into returning the incorrect label. These studies, for instance, usually rely on a prediction score that is fragile to gradient-based attacks. In the context of a more realistic situation where an attacker has very little information about the outputs of a malware detection engine, modest evasion rates are achieved. In this paper, we propose a method using reinforcement learning with DQN and REINFORCE algorithms to challenge two state-of-the-art ML-based detection engines (MalConv & EMBER) and a commercial AV classified by Gartner as a leader AV. Our method combines several actions, modifying a Windows portable execution (PE) file without breaking its functionalities. Our method also identifies which actions perform better and compiles a detailed vulnerability report to help mitigate the evasion. We demonstrate that REINFORCE achieves very good evasion rates even on a commercial AV with limited available information.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 19 of its 35 references that are in the map. First posted on arXiv 24 Mar 2022.

Builds on

19 of 35 references.

References · 19

  1. 2022Generative Adversarial Networkscited by 5,603
  2. 2013Intriguing properties of neural networkscited by 5,114
  3. 2013Playing Atari with Deep Reinforcement Learningcited by 3,211
  4. 2016Asynchronous Methods for Deep Reinforcement Learningcited by 2,218
  5. 2018EMBER: An Open Dataset for Training Static PE Malware Machine Learning Modelscited by 164
  6. 2017Malware Detection by Eating a Whole EXEcited by 114
  7. 2017Generating Adversarial Malware Examples for Black-Box Attacks Based on GANcited by 74
  8. 2018Adversarial Malware Binaries: Evading Deep Learning for Malware Detection in Executablescited by 60
  9. 2020Functionality-preserving Black-box Optimization of Adversarial Windows Malwarecited by 58
  10. 2020SOREL-20M: A Large Scale Benchmark Dataset for Malicious PE Detectioncited by 49
  11. 2020A Survey of Machine Learning Methods and Challenges for Windows Malware Classificationcited by 33
  12. 2017Survey of Machine Learning Techniques for Malware Analysiscited by 31
7 more
  1. 2019Explaining Vulnerabilities of Deep Learning to Adversarial Malware Binariescited by 30
  2. 2018Non-Negative Networks Against Adversarial Attackscited by 20
  3. 2021A Survey on Adversarial Attacks for Malware Analysiscited by 15
  4. 2021Malware Analysis with Artificial Intelligence and a Particular Attention on Results Interpretabilitycited by 9
  5. 2020MalFox: Camouflaged Adversarial Malware Example Generation Based on Conv-GANs Against Black-Box Detectorscited by 7
  6. 2020A survey on practical adversarial examples for malware classifierscited by 5
  7. 2019Malware Evasion Attack and Defensecited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 1; 2023: 3; 2024: 3; 2025: 1; 2026: 2.

10 papers.

  1. 2022AI-based Malware and Ransomware Detection Modelscited by 7
  2. 2023Quantum Machine Learning for Malware Classificationcited by 3
  3. 2023Creating Valid Adversarial Examples of Malwarecited by 3
  4. 2023The Power of MEME: Adversarial Malware Creation with Model-Based Reinforcement Learningcited by 3
  5. 2025Semantic Preprocessing for LLM-based Malware Analysiscited by 3
  6. 2024Reinforcement Learning for an Efficient and Effective Malware Investigation during Cyber Incident Responsecited by 2
  7. 2024EGAN: Evolutional GAN for Ransomware Evasioncited by 0
  8. 2024A Novel Reinforcement Learning Model for Post-Incident Malware Investigationscited by 0
  9. 2026Adversarial Malware Generation in Linux ELF Binaries via Semantic-Preserving Transformationscited by 0
  10. 2026REPLICANT: Learning Policies for Evading and Hardening Malware Detectorscited by 0