Standard Intrusion Detection Systems (IDS) have become increasingly inadequate due to the rampant sophistication of cyberattacks (ranging from stealthy advanced persistent threats, to polymorphic malware). Leading-edge models based on deep learning have recognized the power of learning complex spatial, and temporal, patterns from inputs and can learn anomalous or evasive threats, but they are vulnerable to adversarial examples (inputs that are slightly modified to cause misclassifications). This work introduces an adversarially-resilient deep learning framework for detecting evasive (or stealthy) cyber threats in the context of network traffic. The framework blends One-Dimensional Convolutional Neural Networks (1D- CNNs) with Long Short-Term Memory (LSTM) layers in the same architecture to model spatial correlations along with the temporal sequence of flow-based network features. Efficient feature selection is performed using Recursive Feature Elimination (RFE) to eliminate a large amount of noise and dimensionality. Finally, adversarial training utilizing the Fast Gradient Sign Method (FGSM) is incorporated to create ferry adversaries to the model. The construct is trained and tested with CICIDS2017 to validate the methodology, benchmarking a number of modern day attack scenarios in the process. Experiment-based results reveal the superior performance of the proposed framework, achieving a clean test data F1-score of 91.7%, and a robust performance above the 80% threshold with an F1-score of 84.5% under FGSM-based attacks. Tests comparing the proposed model, and its robustness and accuracy under similar attacks against the baseline models Support Vector Machine (SVM), Random Forest and vanilla Multilayer Perceptron (MLP) substantiate the advantage of the system we proposed on both points Robustness, and Accuracy. These results suggest the use of adversarially trained hybrid deep learning models for intrusion detection in real-world applications is worth pursuing, providing an effective and adaptable way to secure critical infrastructure in the digital era.

[1] Alarfaj F. and Khan N., “Enhancing the Performance of SQL Injection Attack Detection Through Probabilistic Neural Networks,” Applied Sciences, vol. 13, no. 7, pp. 1-11, 2023. DOI: 10.3390/app13074365

[2] Alqarni A., Alsharif N., Khan N., Georgieva L, and et al, “MNN-XSS: Modular Neural Network Based Approach for XSS Attack Detection,” Computers, Materials and Continua, vol. 70, no. 2, pp. 4075-4085, 2022. DOI: 10.32604/cmc.2022.020389

[3] Alshehri A., Khan N., Alowayr A., and Alghamdi M., “Cyberattack Detection Framework Using Machine Learning and User Behavior Analytics,” Computer Systems Science and Engineering, vol. 44, no. 2, pp. 1679-1689, 2023. DOI: 10.32604/csse.2023.026526

[4] Al-Shurbaji T., Anbar M., Manickam S., Hasbullah I., and et al., “Deep Learning-Based Intrusion Detection System for Detecting IoT Botnet Attacks: A Review,” IEEE Access, vol. 13, pp. 11792-11822, 2025. DOI: 10.1109/ACCESS.2025.3526711

[5] Azam Z., Islam M., and Huda M., “Comparative Analysis of Intrusion Detection Systems and Machine Learning-Based Model Analysis Through Decision Tree,” IEEE Access, vol. 11, pp. 80348-80391, 2023. DOI: 10.1109/ACCESS.2023.3296444

[6] Bansal K. and Singhrova A., “Enhanced IIoT security: A Hybrid Intrusion Detection Framework Using Signature-Based and Deep Learning Approaches,” The International Arab Journal of Information Technology, vol. 23, no. 2, pp. 233-245, 2026. DOI: 10.34028/iajit/23/2/5

[7] Canadian Institute for Cybersecurity, Intrusion Detection Evaluation Dataset (CIC-IDS2017), https://www.unb.ca/cic/datasets/ids-2017.html, Last Visited, 2025.

[8] El-Amir S., “Comprehensive Cybersecurity Review: Modern Threats and Innovative Defense Approaches,” International Journal of Computers and Informatics (Zagazig University), vol. 1, pp. 30-37, 2023. https://www.ijci.zu.edu.eg/index.php/ijci/article/v iew/77

[9] Khalaf L., Alhamadani B., Ismael O., Radhi A., and et al., “Deep Learning-Based Anomaly Detection in Network Traffic for Cyber Threat Identification,” in Proceeding of the 24th Cognitive Models and Artificial Intelligence Conference., Istanbul, pp. 303-309, 2024, DOI: 10.1145/3660853.3660932

[10] Khan N., Abdullah J., and Khan A., “Defending Malicious Script Attacks Using Machine Learning Classifiers,” Wireless Communications and Mobile Computing, vol. 2017, pp. 1-9, 2017. DOI: 10.1155/2017/5360472

[11] Li B., Wu Y., Song J., Lu R., and et al., “DeepFed: Federated Deep Learning for Intrusion Detection in Industrial Cyber-Physical Systems,” IEEE Transactions on Industrial Informatics, vol. 17, no. 8, pp. 5615–5624, 2021. DOI: 10.1109/TII.2020.3023430

[12] Liu Q., Hagenmeyer V., and Keller H., “A Review of Rule Learning-Based Intrusion Detection Systems and their Prospects in Smart Grids,” IEEE Access, vol. 9, pp. 57542-57564, 2021. DOI: 10.1109/ACCESS.2021.3071263

[13] Mengara A., Yoo Y., and Leung V., “IoTSecUT: Uncertainty-Based Hybrid Deep Learning Approach for Superior IoT Security Amidst Evolving Cyber Threats,” IEEE Internet of Things Journal, vol. 11, no. 16, pp. 27715-27731, 2024. https://ieeexplore.ieee.org/document/10538199

[14] Sagar R., Jhaveri R., and Borrego C., “Applications in Security and Evasions in Machine Learning: A Survey,” Electronics, vol. 9, no. 1, pp. 1-42, 2020. DOI: 10.3390/electronics9010097

[15] Sanikommu V., Mummaneni S., Jacob N., Emmanuel K., and Variam R., “Deep Learning- Based Control System for Context-Aware Surveillance Using Skeleton Sequences from IP and Drone Camera Video,” The International Arab Journal of Information Technology, vol. 22, no. 5, pp. 919-929, 2025. DOI: 10.34028/iajit/22/5/6

[16] Song H. and Kim H., “Self-Supervised Anomaly Detection for In-Vehicle Network Using Noised Pseudo Normal Data,” IEEE Transactions on Vehicular Technology, vol. 70, no. 2, pp. 1098- 1108, 2021. DOI: 10.1109/TVT.2021.3051026

[17] Wong J., A. M. Berenbeim, Bierbrauer D., and Bastian N., “Uncertainty-Quantified, Robust Deep Learning for Network Intrusion Detection,” in Proceeding of the. Winter Simulation Conference. (WSC), Texas, pp. 2470-2481, 2023. DOI: 10.1109/WSC60868.2023.10407559

[18] Yang J. and Lim H., “Deep Learning Approach for Detecting Malicious Activities over Encrypted Secure Channels,” IEEE Access, vol. 9, pp. 39229- 39244, 2021. DOI: 10.1109/ACCESS.2021.3064561