IOT and Machine Learning Based Real-Time Protection Against Cyber-Physical Threat Mitigation
Harnit Saini, Faz Mohammad, Manoj Kumar Singh, Kapil Sharma, Meenu Shukla, Ramendra Pratap Singh · 2024
An investigation is carried out into the dynamic relationship between machine learning algorithms used in IoT security, with detailed scrutiny of performance indicators to give a fuller view of what happens. The study pays particular attention to metrics like accuracy, precision, recall, and F1 score when ten varied samples are being examined. As a result, all the algorithms exhibit excellent performance with $95.5 \%-98.2 \%$ accuracy values, inspiring high reliability under diverse scenarios. Precision and recall metrics also depict the models’ capabilities. Balancing precision and recall can be observed in F1 scores that are between $95.2 \%$ and $98.0 \%$, and these numbers show that the approach based on this understanding of trade-offs is characterized by its operationalization. At the other end, results obtained from performance metrics go beyond numerical benchmarks to give a holistic appreciation of the algorithms’ strength and areas for possible enhancement. The paper also does not provide only a theoretical approach to machine learning related to security in the IoT environment but also provides information about security considerations for the successful deployment of powerful, adaptive models into real-life scenarios. In this case, findings of the research help in understanding how to make sure that smart and effective measures are implemented during the process of developing security solutions considering all related issues in connection with the IoT technology landscape.