Threat Analysis and Prediction in IoT Devices Using Machine Learning
KONTAGORA KONTAGORA, Muhammad Mamman, ADESHINA ADESHINA, Addicott Steve, HABIBA HABIBA, Musa Musa · International Journal of Research Publication and Reviews · 2025
The rapid adoption of Internet of Things (IoT) devices across various domains, including healthcare, smart cities, and industrial automation, has introduced significant security challenges.IoT devices are inherently vulnerable due to limited resources, reliance on legacy protocols, and lack of standardized security frameworks.This study investigates the application of machine learning models for threat analysis and prediction in IoT environments.Using the CICIoT2023 dataset, which comprises diverse IoT network traffic data, three machine learning models-Random Forest (RF), Support Vector Machine (SVM), and Deep Neural Networks (DNN)-were evaluated for their performance in detecting and mitigating security threats.The results demonstrate that RF outperforms the other models with an accuracy of 99.15%, precision of 99%, recall of 99%, and an F1-score of 99.06%, making it the most suitable model for real-time IoT threat detection.DNN achieved high accuracy (98.18%) but was limited by its computational demands, while SVM lagged significantly with an accuracy of 83%.Feature analysis identified packet size, protocol types, and connection duration as critical predictors of malicious activity.To address resource constraints in IoT devices, an API was developed for integrating the models into IoT gateways, enabling real-time deployment.This study highlights the potential of RF in enhancing IoT security and underscores the need for optimizing ML models for resource-limited environments.Future work should focus on hybrid models, edge computing, and real-world validation to further advance IoT security solutions.These findings contribute to the development of scalable and efficient intrusion detection systems for IoT ecosystems.