Comparative Analysis of Conventional and Hybrid Machine Learning Model for Intrusion Detection Systems

Due to the rapid expansion of the IoT ecosystem, standard intrusion detection system (IDS) fails in complex, resource-constrained IoT environments, posing security vulnerabilities. Traditional security systems often fall short to address the dynamic nature of IoT environment and can't handle large amount of data produced by IoT devices over network resulting in low detection accuracy, processing delays, and false positives. To overcome this limitation, there is a need to develop hybrid machine learning-based IDS that uses several machine learning (ML) models to improve IoT intrusion detection efficiency, accuracy, and adaptability. Combining supervised and unsupervised learning under the IDS paradigm might help to enhance attack classifications and detection rates. Feature selection and dimensionality reduction may use to enhance data processing, hence lowering computer overhead without compromising necessary characteristics to differentiate between safe and risky activity. This work contrasts and evaluates the conventional machine learning models and hybrid machine learning model using key performance metrics like accuracy, precision, recall, F1 Score. The hybrid model offers promising solution as compared to conventional machine learning models such Logistic Regression, Random Forest, and Naive Bayes in terms of attack detection accuracy and false positive rates. Based on experimental results, it has been found that Hybrid model yields high training accuracy of 96%, precision of 0.95, recall of 0.97 and F1-score of 0.96 that is high as compared to conventional machine learning models. Random forest yields high training accuracy of 94% that is more than naive bayes 87% and logistic regression 89%.