An Intelligent IoT Attack Detection Model Using Weighted ELM and Educational Achievement Guided Optimization
Chandrakanth Reddy Borra, Srinivas Cheekati, Piyush Kumar Pareek, Ramya Vani Rayala, Sri Raman Kothuri, V. Kumar Chinnaiyan · 2025
The Internet of Things (IoT) connects smart gadgets all around the world so they may talk to each other and share data automatically. The IoT allows for massive data collection and analysis, which improves quality of life in many areas. IoT data in particular contains a wealth of information useful for detecting anomalies. Cybersecurity faces both opportunities and challenges from the heterogeneous nature of the IoT. For datasets with heterogeneous properties, the traditional methods of cybersecurity monitoring may not work since they necessitate alternative types of data pre-processing and management. Anomaly detection benefits greatly from the different signal sets captured by heterogeneous types of network devices, as opposed to readings from a single type of device. The authors of this paper offer a solution to these problems by combining EAGTOA with Weighted Extreme Learning Machine (W-ELM) to create an improved attack detection model. In contrast to EAGTOA, which optimizes feature selection and learning by mimicking an educational learning environment wherein network parameters are modified iteratively, the W-ELM model improves classification effectiveness through dynamically altering network weights. An extensively used benchmark dataset for network intrusion detection research, the NSL-KDD, is utilized to test the suggested technique. The W-ELM-EAGTOA framework achieves better outcomes than traditional machine learning models in terms of recall, accuracy, precision, and false positive rate, according to the experimental data. Better generalizability and flexibility to differences in IoT network traffic are guaranteed by combining weight-adjusted learning with guided optimization. In conclusion, this study provides a smart and computationally efficient approach to detecting attacks on IoT networks, which improves both the scalability and the accuracy of such detections. To further enhance IoT security frameworks, future research will center on realtime implementation, resistance to adversarial attacks, and hybrid integration with deep learning models.