Chukumeka Gift Iroanwusi, Friday Eleonu Onuodu, Davies Isobo Nelson, Bassey Aniefiok Tom
Abstract
The growing sophistication of cyber threats demands intelligent, adaptive security solutions capable of detecting malicious network activity in real time. This study presents a hybrid cybercrime prediction model that integrates Artificial Neural Networks (ANN), Autoencoders, Support Vector Machines (SVM), and XGBoost for the classification of network traffic as either benign or malicious. Structured network traffic features extracted from the CIC-DDoS2019 dataset including IP addresses, ports, timestamps, protocol types, and payload sizes were used to train and evaluate the model using a 70:30 train-test split. To address privacy concerns in network data analysis, the system incorporates the CKKS homomorphic encryption scheme, enabling secure computation on encrypted data without exposing sensitive information. Additionally, the system employs heuristic URL analysis integrated with Google Safe Browsing and VirusTotal APIs for phishing site detection. Experimental results demonstrate that the system accurately classifies both DDoS and benign traffic patterns, and correctly identifies phishing URLs with a 35% detection rate among tested URLs. Furthermore, the encryption and decryption operations performed within milliseconds confirm the practical efficiency of the system’s privacy mechanism. The proposed framework offers a robust, privacy-preserving approach to cybercrime prediction, combining predictive accuracy with data confidentiality for real-world deployment.
This study presented a hybrid cybercrime prediction system that integrates ANN, Autoencoder, SVM, and XGBoost for network traffic classification, alongside a phishing URL detection module and a CKKS-based homomorphic encryption mechanism for data privacy. By combining multiple machine learning paradigms within a single framework, the system achieves greater predictive robustness than single-algorithm approaches, addressing both the detection accuracy and adaptability challenges inherent in modern cybercrime scenarios.
The experimental results confirm that the system successfully classifies DDoS and benign traffic patterns from the CIC-DDoS2019 dataset, detects phishing URLs with high precision using multi-API validation, and performs encryption and decryption operations with negligible latency (under 5 ms), making it practically deployable in real-time environments. The identification of Port 23 (Telnet) as a primary attack vector further contributes actionable insights for network security practitioners.
The incorporation of homomorphic encryption into the detection pipeline represents a meaningful step towards privacy-aware cybersecurity, ensuring that sensitive network data is never exposed during analysis. This addresses a gap in many existing systems where detection performance is prioritized at the expense of data confidentiality.
Future work should focus on evaluating the system at scale using the full CIC-DDoS2019 benchmark, incorporating larger phishing datasets, and testing against adversarial attack patterns. Extending the framework to support additional attack categories including ransomware, insider threats, and zero-day exploits would further enhance its applicability across diverse cybersecurity contexts. The integration of federated learning could also be explored to enable distributed, privacy-preserving model training across multiple organizational nodes.
References
[1] A. F. Odesanmi, O. A. Ibitoye, O. Iwelumor, H. Obuene, O. T. Arowolo, I. D. Olusegun, et al., “The double-edged sword of digital connectivity: Advancing society and empowering cybercriminals,” Ilorin Journal of Education, vol. 46, pp. 135-147, 2025.
[2] M. Abdullah, M. M. Nawaz, B. Saleem, M. Zahra, E. binte Ashfaq, and Z. Muhammad, “Analytics-Driven Insights into Cybercrime Evolution, Trends, and Defense Strategies: A Comprehensive Survey,” 2025.
[3] J. Manikandan, P. Hemalatha, K. Jayashree, and P. Rajeswari, “Navigating the Digital Landscape: Understanding, Detecting, and Mitigating Cyber Threats in an Evolving Technological Era,” Securing Cyber‐Physical Systems: Fundamentals, Applications and Challenges, pp. 199-223, 2026.
[4] O. Afolalu and M. S. Tsoeu, “Artificial Intelligence as the Next Frontier in Cyber Defense: Opportunities and Risks,” Electronics, vol. 14, p. 4853, 2025.
[5] E. N. Chukwuani, O. R. Odunsi, and C. D. Ikemefuna, “Machine learning techniques for real-time malware classification and threat detection in distributed systems,” World Journal of Advanced Research and Reviews, vol. 26, pp. 2378-2398, 2025.
[6] P. Waghmode, M. Kanumuri, H. El-Ocla, and T. Boyle, “Intrusion detection system based on machine learning using least square support vector machine,” Scientific Reports, vol. 15, p. 12066, 2025.
[7] L. Elluri, V. Mandalapu, P. Vyas, and N. Roy, “Recent advancements in machine learning for cybercrime prediction,” Journal of Computer Information Systems, vol. 65, pp. 249-263, 2025.
[8] A. Shanmugam, “A Comparative Analysis of Kernel-Based Support Vector Machines (SVM) and Convolutional Neural Networks (CNN) for zero-day Malware Detection,” Dublin, National College of Ireland, 2025.
[9] F. A. Vadhil, M. L. Salihi, and M. F. Nanne, “Machine learning-based intrusion detection system for detecting web attacks,” IAES International Journal of Artificial Intelligence, vol. 13, pp. 711-721, 2024.
[10] M. A. Alsoufi, M. M. Siraj, F. A. Ghaleb, A. H. Abdulqader, E. Ali, and M. Omar, “An anomaly intrusion detection systems in iot based on autoencoder: A review,” in International Conference of Reliable Information and Communication Technology, 2023, pp. 224-239.
[11] M. Hesham, M. Essam, M. Bahaa, A. Mohamed, M. Gomaa, M. Hany, et al., “Evaluating predictive models in cybersecurity: A comparative analysis of machine and deep learning techniques for threat detection,” in 2024 Intelligent Methods, Systems, and Applications (IMSA), 2024, pp. 33-38.
[12] İ. Avcı and M. Koca, “Cybersecurity attack detection model, using machine learning techniques,” Acta Polytechnica Hungarica, vol. 20, pp. 29-44, 2023.
[13] O. E. Taylor and I. N. Davies, “A Model for Enhancing Security and Privacy in Pervasive Computing using Homomorphic Encryption ” International Journal of Computer Sciences and Engineering, vol. 13, pp. 21-29, 2025.
[14] R. Alshaya and S. E. Khediri, “Optimizing cybercrime detection: A hybrid deep learning approach for enhanced intrusion detection systems,” Peer-to-Peer Networking and Applications, vol. 18, p. 145, 2025.
[15] I. N. Davies, O. E. Taylor, V. I. E. Anireh, and E. O. Bennett, “Adaptive Hybrid Case-Based Neuro-Fuzzy Model for Intrusion Detection and Prevention for Smart Home Network,” International Journal of Computer Sciences and Engineering, vol. 10, pp. 01-10, 2024.
How to Cite This Paper
Chukumeka Gift Iroanwusi, Friday Eleonu Onuodu, Davies Isobo Nelson, Bassey Aniefiok Tom (2026). Development of Cybercrime Prediction Model using Hybridized Algorithms. International Journal of Computer Techniques, 13(2). ISSN: 2394-2231.
