Machine Learning-Based Cybercrime Detection and Secure Wireless Communication Frameworks for Enhancing Data Privacy, Network Security, and Smart Technology Infrastructure

Abstract

In the Internet of Medical Things (IoMT) context, cybercrime detection is critical since connected medical devices are responsible for collecting, processing, and transmitting sensitive patient information via vulnerable healthcare communication networks. These smart healthcare environments face cyber threats like spoofing, man-in-the-middle attacks, data injection, unauthorized access and service disruption. This study proposes and tests a cybercrime detection, intrusion detection, and privacy protection framework based on machine learning with the WUSTL-EHMS-2020 dataset. The dataset consists of 16,318 records with 44 features, such as network-flow features, biometric attributes, and a binary target label (normal and attack traffic). Six supervised machine learning models were tested after preprocessing, feature scaling and stratified 80:20 train-test splitting using accuracy, precision, recall, F1-score, ROC-AUC, false positive rate, false negative rate and confusion matrix. The results indicate that the KNN algorithm has the best balanced performance with 95.50% accuracy, 88.30% precision, 73.84% recall, 80.43% F1-score, and 92.34% ROC-AUC. Gradient Boosting had the best ROC-AUC score of 94.16%, which means that it was good in separating the classes. Through feature analysis, it was found that the variables of network traffic were the most significant features in detecting attacks. The proposed framework will help ensure patient data privacy, secure healthcare communication, network security, and resilient smart technology infrastructure.