XAInomaly: Explainable, Interpretable and Trustworthy AI for xURLLC in 6G Open-RAN

Artificial intelligence (AI) has already been incorpo-rated into wide range applications of the fifth generation (5G) networks. The AI-native design of 6G network is serving as cornerstone for intelligent, autonomous, and dynamic network operations. AI-driven techniques, such as machine learning (ML) and Deep Learning (DL), facilitate real-time data analytics, predictive modeling, and decision-making processes to optimize resource utilization, enhance network performance, and ensure seamless connectivity for a multitude of devices and services. However, it is crucial in many respects that these AI algorithms are reliable, trustworthy, and explainable. In this direction, Explainable AI (XAI) will ensure transparent and secure operation at different layers of 6G networks. With the integration of XAI, 6G networks can achieve transparent dynamic self-configuration, self-optimization, and self-healing capabilities, enabling the network to adapt to fluctuating demands, mitigate potential issues proactively. To ensure that the AIML algorithms used in 6G Next-generation URLLC (xURLLC) use case are trustable and reliable, we proposed a XAInomaly framework that use our novel fastSHap-cXai method which handle real-time XAI layer operations on Open-RAN (O-RAN). Our performance results show that fastSHAP-C provides a 25% advance over its competitors in terms of resource utilization.