AI Powered Threat Detection in Cybercrime Using Supervised ML Models

The increasing sophistication of cybercrime is a big threat to digital infrastructure, and therefore smart and efficient detection mechanism is required. The presented research dwells upon identifying security threats using AI-driven supervised learning practices. It is mainly aimed at enhancing accuracy and seamlessness of cyber threat detection based on applying machine learning models that have been trained using past information that has been based on patterns. To do that, several supervised models were used, comprising Naive Bayes, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN) and Decision Tree. Every algorithm was trained and tested on annotated set of the features of cybercrimes. Results show Random Forest algorithm has achieved highest accuracy of 99.73 percentage levels as well as F1-score of 0.91 greater than Decision Tree model with accuracy and F1-score of 99.31 percentage levels and 0.84, respectively. Models like Naive Bayes and Support Vector Machine had all performed poorly, but other models like KNN fetched acceptable results. The Random Forest algorithm demonstrated sufficiently balanced results in different measures of evaluation, which reflected the potential to be used in practical purposes, where precision and computational performance have to be considered. The research is a contribution to an increasing field of AI-supported cybersecurity by replicating comparative side-by-side comparisons between learning algorithms paired to detect cybercrime actions. The results indicate the efficiency of such ensemble algorithms as Random Forest and Decision Tree in identifying threats with the high recall and precision. These insights will be useful to the cybersecurity experts as they endeavor to seek appropriate models to build smart threat monitoring systems that remain adaptive to the changing way of attacks in cyberspace.