HGLV-DNN: An Efficient Cyber-Attack Detection Using Hybrid Global-Local-Visual Deep Neural Network

Asmaa Rashid Salman; Noor Muneam Abbas; Duaa H. Hameed; Nidaa Flaih Hassan

doi:10.29284/t2q7em86

Authors

Asmaa Rashid Salman Computer Science College, University of Technology, Baghdad, Iraq
Noor Muneam Abbas Computer Science College, University of Technology, Baghdad, Iraq
Duaa H. Hameed Computer Science College, University of Technology, Baghdad, Iraq
Nidaa Flaih Hassan Computer Science College, University of Technology, Baghdad, Iraq

DOI:

https://doi.org/10.29284/t2q7em86

Keywords:

.

Abstract

This study proposes HGLV-DNN, a hybrid global–local–visual deep neural framework for accurate and robust multi-class network intrusion detection. The method first converts preprocessed CIC-IDS2017 network flows into time–frequency scalogram images via the Continuous Wavelet Transform to capture transient and multiscale patterns. These images are processed in parallel by a Convolutional Neural Network (for local spatial features) and a Transformer (for long-range global dependencies). The resulting feature sets are concatenated, reduced with Principal Component Analysis, and classified using a Random Forest ensemble to enhance stability and generalization. The pipeline includes targeted undersampling and oversampling augmentation to produce a balanced dataset (75% for training and 25% for testing) and applies 10-fold cross-validation during evaluation. Experimental results on the CIC-IDS2017 dataset demonstrate near-perfect detection performance with an Area Under the ROC Curve (AUC) of 0.9999 and overall accuracy of 99.92%, along with Precision = 99.91%, Recall = 99.90%, and F1-score = 99.90%. The proposed HGLV-DNN framework outperforms several baseline and hybrid approaches, with the closest competitor (Decision Tree) achieving 99.81% accuracy. These findings indicate that integrating CWT-based visual representations with complementary local and global deep feature extractors and ensemble classification yields a highly effective and noise-resilient intrusion detection solution.

References

[1] Abdulganiyu, O. H., Ait Tchakoucht, T., & Saheed, Y. K. (2023). A systematic literature review for network intrusion detection system (IDS). International journal of information security, 22(5), 1125-1162.

[2] Al-Safaar, D. W., & Al-Yaseen, W. L. (2023). A survey of network intrusion detection systems based on deep learning approaches. Научно-технический вестник информационных технологий, механики и оптики, 23(2), 352-363.

[3] Kimanzi, R., Kimanga, P., Cherori, D., & Gikunda, P. K. (2024). Deep Learning Algorithms Used in Intrusion Detection Systems--A Review. arXiv preprint arXiv:2402.17020.

[4] AbdelHalim, A. P., Abo-Alian, A., & Badr, N. (2025). Deep Learning Techniques for Network Intrusion Detection: A Comparative Survey. International Journal of Intelligent Computing and Information Sciences, 25(2), 74-87.

[5] Al Lail, M., Garcia, A., & Olivo, S. (2023). Machine learning for network intrusion detection—a comparative study. Future Internet, 15(7), 243.

[6] Gueriani, A., Kheddar, H., & Mazari, A. C. (2023, November). Deep reinforcement learning for intrusion detection in IoT: A survey. In 2023 2nd International Conference on Electronics, Energy and Measurement (IC2EM) (Vol. 1, pp. 1-7). IEEE.

[7] Ganapaneni, S. B. A. K. S., & Najana, M. Attention-Based Deep Learning Frameworks for Network Intrusion Detection: An Empirical Study.

[8] Enisoglu, R., & Rakocevic, V. (2025). A Novel Wavelet Transform and Deep Learning-Based Algorithm for Low-Latency Internet Traffic Classification. Algorithms, 18(8), 457.

[9] Zeng, Y. (2025). CSAGC-IDS: A Dual-Module Deep Learning Network Intrusion Detection Model for Complex and Imbalanced Data. arXiv preprint arXiv:2505.14027.

[10] Puri, V., Chandekar, P., & Chandan, S. (2025). Enhancing IoT Security Through a Hybrid Deep Learning Model: CNN Meets Transformer for Robust Intrusion Detection.

[11] Rismawan, V. L., & Pramudya, E. R. (2025). Network Intrusion Detection System Using Convolutional Neural Network and Random Forest Classifiers. INOVTEK Polbeng-Seri Informatika, 10(2), 753-761.

[12] Du, C., Guo, Y., & Zhang, Y. (2024). A Deep Learning-Based Intrusion Detection Model Integrating Convolutional Neural Network and Vision Transformer for Network Traffic Attack in the Internet of Things. Electronics (2079-9292), 13(14).

[13] Xiang, B., Zheng, R., Zhang, K., Li, C., & Zheng, J. (2025). FFT-RDNet: A Time–Frequency-Domain-Based Intrusion Detection Model for IoT Security. Sensors, 25(15), 4584.

[14] Zhang, C., Li, J., Wang, N., & Zhang, D. (2025). Research on Intrusion Detection Method Based on Transformer and CNN-BiLSTM in Internet of Things. Sensors, 25(9), 2725.

[15] Koukoulis, I., Syrigos, I., & Korakis, T. (2025). Self-Supervised Transformer-based Contrastive Learning for Intrusion Detection Systems. arXiv preprint arXiv:2505.08816.

[16] Li, J. (2025). Deep Learning Models in Network Intrusion Detection Systems. Applied Mathematics & Nonlinear Sciences, 10(1).

[17] Sun, Z., An, G., Yang, Y., & Liu, Y. (2024). Optimized machine learning enabled intrusion detection 2 system for internet of medical things. Franklin Open, 6, 100056.

[18] Bukhari, O., Agarwal, P., Koundal, D., & Zafar, S. (2023). Anomaly detection using ensemble techniques for boosting the security of intrusion detection system. Procedia Computer Science, 218, 1003-1013.

[19] Vishwakarma, M., & Kesswani, N. (2023). A new two-phase intrusion detection system with Naïve Bayes machine learning for data classification and elliptic envelop method for anomaly detection. Decision Analytics Journal, 7, 100233.

[20] Rajathi, C., & Rukmani, P. (2025). Hybrid Learning Model for intrusion detection system: A combination of parametric and non-parametric classifiers. Alexandria Engineering Journal, 112, 384-396.

[21] Bibers, I., Arreche, O., Alayed, W., & Abdallah, M. (2025). Ensemble-IDS: An Ensemble Learning Framework for Enhancing AI-Based Network Intrusion Detection Tasks. Applied Sciences, 15(19), 10579.

[22] Khan, Adnan Shahid, Zeeshan Ahmad, Johari Abdullah, and Farhan Ahmad. "A spectrogram image-based network anomaly detection system using deep convolutional neural network." IEEE access 9 (2021): 87079-87093.

[23] Hazim Obaid, Zahraa, Behzad Mirzaei, and Ali Darroudi. "An efficient automatic modulation recognition using time–frequency information based on hybrid deep learning and bagging approach." Knowledge and Information Systems (2024): 1-18.