Securing IoT System Using ML Models
Arbaz Adib Dalwai, Shivani Jaswal, Rohit Kumar Verma · Auerbach Publications eBooks · 2025
The security mechanisms for Internet of Things systems often rely on traditional rule-based detection systems alone to detect sophisticated threats like distributed denial of service, which sometimes fall inadequate in terms of adaptability required in modern-day detection. This research tries to bridge the gap between the academic researches and practical applications by contributing a scalable and robust detection framework in modern cloud infrastructure. The research aims to design and implement a real-time hybrid detection mechanism on a cloud platform—Amazon Web Services that would integrate the security information and event management, intrusion detection system (IDS), and machine learning models to detect and classify the cyber threats efficiently. Attack simulations were conducted to generate real-time logs which were monitored through the Suricata IDS, followed by the log processing in ELK (Elasticsearch, Logstash, and Kibana) stack. Ensemble learning models like Random Forest and XGBoost were deployed to complement the rule-based detections and all this was presented in visual forms in real time without significant delays. This was proven by an average detection time of 0.514 milliseconds, demonstrating the system’s suitability for real-world conditions. The framework tends to bridge the gap between conceptual and practical deployments with the implementation of real-time hybrid detection system in a cloud environment.