Vol. 2, Issue 1, Part A (2025)

The evolution of next-generation network infrastructures, including 5G, IoT, and software-defined environments, has introduced unprecedented complexity in cybersecurity management. Traditional intrusion detection systems (IDS) often fail to detect novel, adaptive, and stealthy threats due to their reliance on static signatures and limited scalability. This study presents the design and evaluation of an AI-enhanced Intrusion Detection System that integrates deep hybrid learning, adversarial robustness, and explainable artificial intelligence (XAI) to overcome these limitations. Utilizing benchmark datasets such as UNSW-NB15, CIC-IDS2017, BoT-IoT, and ToN-IoT, the proposed model combines a Convolutional Neural Network (CNN) and a Bidirectional Long Short-Term Memory (BiLSTM) architecture, reinforced with online learning for real-time adaptability. Experimental results reveal that the system achieved an average F1-score exceeding 0.97 and reduced false-positive rates by up to 60% compared with traditional models like SVM, Random Forest, and LSTM. The incorporation of adversarial training significantly enhanced resilience to evasion attacks, while the online update mechanism allowed rapid recovery from concept drift in streaming data. Moreover, the use of SHAP and LIME frameworks introduced interpretability, enabling analysts to visualize and understand detection decisions without compromising accuracy. Statistical validation confirmed the model’s superiority in detection precision, computational efficiency, and robustness under realistic high-throughput conditions. The findings underscore the critical role of AI-driven approaches in establishing adaptive, interpretable, and resilient IDS framework, ensuring scalability and interpretabilitys suited for dynamic network ecosystems. The study concludes that deploying AI-augmented IDS solutions with integrated learning, transparency, and robustness mechanisms is essential for securing future digital infrastructures. It also recommends that future research focus on large-scale real-time implementations across 5G slices and industrial IoT environments to validate operational scalability and policy compliance.