A Trust-By-Learning Framework for Secure 6G Wireless Networks Under Native Generative AI Attacks

Sixth-generation (6G) wireless networks will become vulnerable due to native generative AI (GenAI)-driven intelligent poisoning attacks in both the radio unit and the core network. In particular, network parameters and metrics in cross-layer design pose fundamentally uncertain conditions and can be compromised through the native GenAI mechanism, which leverages data augmentation and reconstruction capabilities. This work investigates the capabilities of native GenAI to create novel poisoning attacks in wireless networks, while investigating their impact through uncertainty-informed root analysis. Then, detected attacks are mitigated by developing a trustworthy service aggregation in the wireless network. First, a joint decision problem is formulated to generate intelligent poisoning attacks, understand their root cause by defining a new measure of uncertainty as plausibility, and mitigate them through trustworthy service aggregation in wireless networks. Second, to address the challenges of the formulated problem, a novel Trust-By-Learning (TBL) framework is developed. The proposed TBL framework primarily consists of three components: 1) a native GenAI mechanism that can penetrate intelligent poisoning attacks in wireless networks’ metrics and parameters; 2) a Dempster-Shafer-based evidence theoretic mechanism that is developed to understand the root cause of inherently uncertainty of those attacks to quantify the trust for further mitigation; and 3) a meta-reinforcement-based Markov Decision Process learning framework that can mitigate the intelligent attacks by enforcing trustworthy service aggregation. Extensive experimental analysis demonstrates that native GenAI methods, such as generative adversarial network (GAN), variational autoencoder (VAE), and autoencoder have significant capability to enforce poisoning attacks. Results show that the autoencoder performs significantly better in generating poisoning attacks capabilities of 98.2%, 97.4%, and 95% for Amazon, Netflix, and Download services, respectively. The proposed TBL framework effectively replicates intelligent attack dependencies by achieving a trust score of 0.972, 0.922, and 0.892 for Amazon, Download, and Netflix services, respectively. Finally, the proposed TBL framework shows efficacy in understanding the trust in GenAI-driven intelligent poisoning attacks on network parameters and metrics by quantifying root causes and mitigating rates.