نوع مقاله : علمی-پژوهشی
نویسنده
عضو هیات علمی / دانشگاه سجاد
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسنده [English]
With the rapid increase in the number and diversity of malware, traditional signature-based methods face limitations in detecting new and unknown samples. The constantly evolving structure of malware and the use of sophisticated obfuscation techniques have created a need for intelligent methods based on machine learning and deep learning. In recent years, machine learning-based methods have demonstrated effective malware detection performance by extracting behavioral patterns and distinguishing features. However, many of these methods rely on feature engineering, and their performance depends heavily on the correct selection of features. In this study, to reduce reliance on manual feature extraction, a deep convolutional neural network (CNN) is employed to automatically learn effective features. Subsequently, to enhance decision-making robustness, the extracted features are fed into multiple independent classifiers, and their outputs are combined using a hard-voting ensemble learning method. The final decision is reached through this ensemble approach; specifically, if at least three classifiers identify a sample as malware, it is ultimately labeled as "malware." Experimental results from prior research indicate that hybrid models combining CNNs with diverse classifiers outperform single models. The results of this study demonstrate a significant improvement in the effectiveness of the proposed approach compared to other models. The findings suggest that the proposed model is an efficient and reliable method for detecting emerging malware in real-world environments.
کلیدواژهها [English]