International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 11 | Nov 2024
p-ISSN: 2395-0072
www.irjet.net
Malware Classification Using Deep Learning Ankit Das1, Sujal Jadhav2, Mohd Saad Khan3, Dhruv Patel4, Keshav Halder5 1 Student, Dept. of Computer Science Engineering (Cyber Security), Thakur College of Engineering and Technology,
Mumbai, Maharashtra
2Student, Dept. of Computer Science Engineering (Cyber Security), Thakur College of Engineering and Technology,
Mumbai, Maharashtra
3Student, Dept. of Computer Science Engineering (Cyber Security), Thakur College of Engineering and Technology,
Mumbai, Maharashtra
4Student, Dept. of Computer Science Engineering (Cyber Security), Thakur College of Engineering and Technology,
Mumbai, Maharashtra
5Student, Dept. of Computer Science Engineering (Cyber Security), Thakur College of Engineering and Technology,
Mumbai, Maharashtra ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Malware attacks are increasing and causing lots
to enhance malware detection and classification. [2] [1] [3]. The research done in recent times has increased its focus on using deep learning models to classify malware into specific families [1] [2] [3]. We will be using converting malware binaries into visual representations, image-based malware classification is a novel technique that enables efficient analysis and classification using deep learning models, especially Convolutional Neural Networks (CNNs). Deep learning models can remain relevant in the face of quickly changing threats because of their capacity to continuously learn from fresh data. Deep learning models are appropriate for real-time detection applications because, once trained, they can categorize new samples rapidly [7] [8]. The shortcomings of conventional malware detection methods, which frequently rely on static or dynamic analysis, are addressed by this approach. By checking for the unique qualities and patterns in malware samples, these models aim to precisely determine which particular malware belongs to which family. This research project is going to compare multiple deep learning models to solve the challenge of malware family categorization. This project can help develop new defenses against emerging malware and enable rapid responses to evolving malware attacks.
of harm to both organizations and normal users. In this study we have categorized malware according to the family to which it belongs. We evaluated the efficacy of six deep learning models—CNN, VGG16, VGG19, RestNet50, Xception and MobileNet—in classifying malware by examining Malimg dataset and examining classes of Malware. MobileNet outperformed other models according to our results, with the maximum accuracy of 98.74%. This illustrates how sophisticated deep learning techniques may be used to improve the classification of malware and emphasizes the necessity of ongoing model adaptation and improvement in order to combat changing cyberthreats. Key Words: Malware, Cybersecurity, AI, Deep Learning, CNN, VGG16, VGG19, MobileNet, Xception, ResNet50, Accuracy, Precision.
1.INTRODUCTION Malware has become less of a novelty and more of a reality for billions of people and organizations around the world in recent years [4]. Malware attacks rose by 11 % to 13,44,566 in 2024 from 12,13,528 in 2023 [5]. Any software that damages, disrupts, gains access to, or steals user data and/or money is referred to as malware. Malware can result in a variety of problems, such as losing control over data, identity theft, private life eavesdropping, hardware failure, service denial, and more. Malware attacks target both normal users and commercial organizations. Whether the malware is infecting a personal computer or a business network, the damage it causes varies depending on the malware type and its intention. Among the most well-known forms of malware include trojans, viruses, worms, ransomware, spyware, and adware [4]. Tools such as Virus Total and PEview are used for the detection and classification of malware. Malware creates a threat to the integrity and authenticity of data.
This project will involve: Collecting and preprocessing the malware image data: We would be using a Dataset which contains 25 malware families/classes. Implementing and Training various deep learning models: We have trained this data on CNN, VGG16, VGG19, RestNet50, Xception, Mobilenet. Evaluating model performance using appropriate metrics: The effectiveness of each model will be measured using metrics such as accuracy and loss. Analyzing results to identify the most effective approaches: Depending on the requirements and evaluation metrics, the best suited model would be selected amongst the other model trained.
Researchers have found the potential of deep learning, a powerful branch of machine learning, as a feasible approach
© 2024, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 18