Certified Causal Attribution for Real-Time Attack Forensics in 6G Network Slicing

The DA-GC framework, which stands for Resource-Conditioned Granger Causality with Axiomatic Resource Contention Model, has reached an attribution accuracy of 89.2% within 87 ms for identifying cross-slice attacks in 6G networks. This was validated using a testbed that emulated 15 slices and included 1,100 attack scenarios. Notably, this represents a 7.9 percentage-point enhancement over the leading baseline while operating at 2.7 times lower latency. The framework employs a certified causal attribution method that combines resource-conditioned Granger causality with an axiomatically established Resource Contention Model, effectively mitigating resource-mediated confounding and overcoming the issue of misleading correlations that typical Granger tests fail to differentiate from true causal relationships.