Intelligent Cybercrime Prediction and Control Model Using Hybrid Machine Learning Algorithms for Ddos and Phishing Attack Mitigation

Abstract

The increasing sophistication and frequency of cyberattacks have created significant challenges for protecting network resources and end users, particularly against attacks such as Distributed Denial-of-Service and phishing. Conventional cybersecurity approaches often focus primarily on attack detection and classification, with limited capability for automated and timely threat mitigation. This study developed an intelligent cybercrime prediction and control model using hybrid machine learning techniques to extend cybercrime prediction beyond threat identification to automated control and mitigation. The proposed model integrates Artificial Neural Network for network traffic prediction, Support Vector Machine for traffic classification, and Autoencoder and XGBoost techniques for network control. The model was developed using structured network traffic features extracted from the CIC-DDoS2019 dataset, which was preprocessed and divided into training and testing subsets using a 70:30 ratio. The developed system was evaluated using DDoS and phishing attack scenarios. For DDoS attacks, the control module processed 84 traffic instances, blocking 22 malicious instances, allowing 52 legitimate instances, and quarantining 10 suspicious instances. The control module achieved a 100% control action execution rate, 100% threat response accuracy, 100% detection-to-control success rate, and 100% control decision accuracy. For phishing attacks, the system achieved a 100% malicious URL blocking rate, 100% legitimate access preservation rate, and 0% false blocking rate. The findings demonstrate that integrating hybrid machine learning-based prediction with automated threat control can enhance proactive cyber defense, reduce reliance on manual intervention, and provide effective mitigation across network-level and application- level cyberattack scenarios.

References