Hybrid Deep Learning Model for Memory-Based Malware Detection

Abstract

With the development of technology, attacks on systems are increasing, detection is becoming more difficult. As a result of this situation, the methods used for malware are insufficient, more effective methods are needed. Within the scope of the study, a model that analyzes memory-based malware behavior and makes classification according to the analysis results is proposed. The CIC-MalMem-2022 memory-based malware behavior analysis dataset created by the Canadian Institute for Cybersecurity (CIC) was preferred. In the proposed model, a 1D Convolutional Neural Network (1D CNN) and Bidirectional Long Short-Term Memory (BiLSTM) based model is developed to classify the malware in the dataset. As a result of the model developed for malware classification, an accuracy rate of 99.98% in binary classification and 92.36% in multiple classification was obtained on the proposed dataset. The obtained result demonstrates the power of the proposed dataset and the developed model in malware classification. © 2025 IEEE.