Enhancing IoT Botnet Detection: An Ensemble Approach for Improved Network Security
Pendem Mukhtesh Venkata Sri Sai, K Venkatraman, D. Sasikala · 2023
Accurate detection of botnets is essential in real world applications such as safeguarding critical infrastructure, ensuring the integrity of online financial transactions, preserving user privacy, and maintaining the reliability of communication networks. However, this period has also witnessed a concurrent rise in network-based threats, including Distributed Denial of Service (DDoS) attacks, wide-reaching spam campaigns, substantial data breaches, and the illegal extraction of sensitive information. Drawing insights from empirical studies and real-world data, our research seeks to unfold the complexities of botnet threats and advance effective detection and mitigation strategies, particularly within the context of IoT devices. The integration of Graph Attention Networks (GAT) and Long Short Term Memory (LSTM) represents an advanced approach to decode the complex relationships embedded within network data. The system achieved a Accuracy of 98.4% and a precision of 98.1% when combined with Random Forest classifier. These advances highlight the significance of robust and adaptable security measures as a fundamental shield, preserving digital ecosystems in the face of ever-changing and progressively intri-cate cybersecurity realm especially in the realm of IoT devices.