Deep Learning Approach for Malicious URL Detection using CNN, RNN, LSTM and Bi-LSTM models

Malicious website protection through malware URL detection is very important for cybersecurity. This paper discusses the use of deep learning algorithms in detecting malware URLs by comparing the performance of four models: Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) networks, and Bidirectional LSTM networks. Each model was trained on a IEEE dataset consisting of URLs, and their accuracy levels were recorded. The highest accuracy rate among these models belongs to Bi-LSTM with 99.25% precision while LSTM had 63.75%, CNN 86.5% and RNN scored worst at 53%. According to these findings, deep learning models such as CNNs and Bidirectional LSTMs are effective in detecting malware URLs, which themselves provide a basis for improving cyber security capabilities. Accordingly, this research argues that deep learning models, specifically CNNs and Bidirectional LSTMs are productive in handling malware URL detection processes thereby providing grounds for boosting internet safety mechanisms in this modern era of technology.