Interpretable Memory-behavior Modeling for Malware Detection Using Additive Explainable Learning

Read full text Summarise with VIRAT
Authors: Noora Alallaq, Muhmmad Al-Khiza’ay, Mohanad Sameer Jabbar, Riyadh Rahef Nuiaa Alogaili, Selvakumar Manickam
Cited by: 0
References: 30

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The increasing popularity of memory-resident and behavior-driven malware has demonstrated some fundamental shortcomings of existing malware detection techniques, especially black box deep learning models.Such models may be accurate at high detection rates, but lack transparency so that analysts cannot trust them, their interpretation is constrained, and their use is not applicable in security critical settings.Overcoming such obstacles is of particular concern to memory-based analysis, the sensitive characteristic of which is the behavioral pattern that determines the difference between malicious and benign activity.The proposed paper suggests inherently explainable malware detection model tailored towards memory-behavior analysis.The method converts low level memory characteristics into semantically significant behavioral signals and uses a clear additive learning organization that breaks each prediction down into overt behavioral contributions.The model is tested against only the CIC-MalMem-2022 dataset, which offers fine-grained behavioral features of memory-related processes of the memory-related activity of modern malware and is well-equipped to be analyzed based on explainability.The experimental data indicate that the suggested model demonstrates a competitive level of detection performance, and the accuracy of around 96% and F1-score of over 96 are obtained, and the false alarm rate is lower than 4%.In addition to the results of detection, the model has a high explanation fidelity (around 0.99), high stability in response to small input perturbations, and is also consistent in its explanations of similar behavior.These results suggest that interpretabilitysensitive model design may provide accurate detection without reducing the transparency.These findings help support the practicability of the explainable malware detection model capable of increasing analytical confidence and operational decision-making.

Fields of Study & Subjects

Advanced Malware Detection Techniques Explainable Artificial Intelligence (XAI) Security and Verification in Computing

Publication details

Journal / source
International journal of intelligent engineering and systems
Volume
19
Issue
4
Pages
531–545
Publication date
23 March 2026
Language
EN