ABSTRACT
The objective of this research is to improve the efficiency in identifying malicious traffic by integrating a deep neural network that utilizes a hierarchical attention mechanism. The intention behind this approach is to augment current methods of detecting detrimental activities through utilizing advanced Machine Learning techniques. It is anticipated that by incorporating a deep neural network model with a hierarchical attention, there will be additional improvement in accurately identifying and categorizing suspicious behavior or content within web traffic data. This mechanism allows for better focus on different aspects or elements within each layer of the model's architecture. It enables fine-grained analysis at both the global level (identifying important features across all layers) and local level (focusing on specific features within individual layers). By combining these two techniques together – deep neural networks along with hierarchical attention –the proposed method significantly improves upon traditional approaches used for malicious traffic detection in terms of speed, accuracy, and robustness. The proposed method consists of incorporating a hierarchical attention mechanism into deep neural network architectures. This technique allows the system to focus its attention on specific regions or features within network traffic flows that are likely indicative of suspicious activity while minimizing noise or false positives. By conducting comprehensive experiments and evaluations, this work contributes valuable insights toward advancements in malware detection using Machine Learning approaches. The findings will pave the way for more robust and efficient systems capable of combating modern-day cyber threats.
