Affiliation:
1. Vellore Institute of Technology, Chennai, India
2. Ganpat University, Gujarat, India
Abstract
The increasing use of the internet and digital devices has led to an exponential growth in cyber-attacks, with malware being one of the most prevalent forms of cybercrime. Modern-day malware is becoming more sophisticated and evasive, using various techniques such as obfuscation, encryption, and code injection to evade detection. To combat this problem, this study proposes a new approach for detecting malware using a convolutional autoencoder with kernel density estimation (KDE). This model uses the autoencoder's encoder to compute KDE and combines reconstruction error with KDE for malware detection. Tested on the malimg dataset, it achieves 98.3% accuracy, comparable to other autoencoder models. This study demonstrates the potential of combining convolutional autoencoder with KDE for detecting modern sophisticated malware, evaluated against existing models using accuracy and precision metrics.
Reference22 articles.
1. Deep Autoencoders and Feedforward Networks Based on a New Regularization for Anomaly Detection
2. On the Resilience of Shallow Machine Learning Classification in Image-based Malware Detection
3. De PaolaA. (2018). Malware Detection through Low-level Features and Stacked Denoising Autoencoders. ITASEC.
4. Hardy, W. (2016). DL4MD: A deep learning framework for intelligent malware detection. Proceedings of the International Conference on Data Science (ICDATA). The Steering Committee of The World Congress in Computer Science, Computer Engineering and Applied Computing (WorldComp).
5. A Malware Detection Approach Using Malware Images and Autoencoders