An AI-Resilient Cybersecurity Framework for 5G Telecommunications Networks: Detecting Adversarial Machine Learning, Autonomous Intrusions, and Intelligent Edge Threats in Critical Infrastructure

@article{1722431,
      author = {Ken Mudzingwa, Stewart Munyaradzi Nyamutswa, Admore Tafadzwa Mugwadzi, Panashe Yolanda Dhegwa, Munashe Naphtali Mupa},
      title = {An AI-Resilient Cybersecurity Framework for 5G Telecommunications Networks: Detecting Adversarial Machine Learning, Autonomous Intrusions, and Intelligent Edge Threats in Critical Infrastructure},
      journal = {Iconic Research And Engineering Journals},
      year = {2026},
      volume = {10},
      number = {2},
      pages = {2497-2514},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1722431.pdf},
      abstract = {Fifth-generation telecommunications networks are becoming a strategic layer of critical infrastructure because they connect public safety, financial systems, healthcare, logistics, energy, industrial control systems and intelligent edge services. The same characteristics that make 5G valuable - ultra-low latency, dense device connectivity, software-defined network functions, network slicing, cloud-native service-based architecture and edge computing - also create an expanded attack surface for adversaries using artificial intelligence. This paper develops an AI-resilient cybersecurity framework for 5G telecommunications networks by integrating 5G security literature, adversarial machine learning research, zero-trust architecture, AI risk management and empirical cyber-incident analytics. The empirical component uses the Kaggle Global Cybersecurity Threats (2015-2024) dataset, comprising 3,000 incident records across 10 countries, 7 industries and 6 attack categories, to model sectoral exposure, telecommunications-specific risk, loss severity and incident-resolution burden. The sample records USD 151.48 billion in aggregate estimated financial loss, 1.51 billion affected-user records and a mean resolution time of 36.48 hours. Telecommunications incidents account for 403 records and USD 20,459.09 million in estimated loss, with man-in-the-middle, DDoS, phishing and malware attacks showing elevated relevance for AI-enabled 5G threat scenarios. Heat-map analysis identifies telecommunications man-in-the-middle attacks as one of the highest AI-5G exposure cells, while random-forest triage demonstrates that affected-user scale, resolution time, year and the constructed AI-5G Exposure Index are the strongest predictors of high-impact incidents in the analytical design. The paper contributes a practical AI-Resilient 5G Cybersecurity Framework built around adversarially robust intrusion detection, zero-trust identity, slice isolation, secure edge orchestration, model-risk governance, threat-informed vulnerability prioritisation, incident-response automation, post-quantum crypto-agility and continuous assurance. The central finding is that 5G security cannot be reduced to conventional perimeter defence; it requires a converged operating model that secures networks, data, machine-learning pipelines, identities, cloud-native functions and critical service continuity together.},
      keywords = {5G security; telecommunications; adversarial machine learning; AI-enabled cyber threats; critical infrastructure; intelligent edge; network slicing; zero trust; intrusion detection; post-quantum cryptography.},
      month = {August},
      doi = {https://doi.org/10.64388/IREV10I2-1722431}
  }