Real-Time Botnet Detection Using Temporal Convolutional Networks and Self-Organizing Map
Amanullah Quamer, Hardarshan Kaur · 2025
Utilizing Internet of Things (IoT) devices has resulted in an increased risk of botnet attacks, posing significant challenges to security. This paper shows a new model that uses a neural network with unsupervised learning, explained using the concept of SHAP. The proposed model achieves training accuracy of 99.97% and validation accuracy 98.31% over 16 epochs. The existing models are compared with the proposed model that focused on the strengths and gaps of each method, particularly showing scalability, real-time detection, and strength to tackle evolving botnet attacks. To analyse this model in real world, we should try to deploy the system in a live network environment using real-time data technologies software (e.g., Apache Kafka). This system will allow us to evaluate the model's ability to process continuous data flows and maintain detection accuracy in dynamic settings.