Birdcast: Interest-aware BEV Multicasting for Infrastructure-assisted Collaborative Perception

This collection highlights emerging security-relevant research across networking and ML-driven control systems. Key findings: (1) Adversarial jamming against DRL-based RAN slicing can induce severe, slice-specific SLA violations and long recovery periods, showing ML-driven control planes are manipulable by budget-constrained attackers. (2) PHY-aware measurements and POLARIS (O-RAN-based) steer UEs to new spectrum with much lower disruption; steering mechanism heterogeneity (BWP vs CA vs HO, etc.) exposes different latency/attack surfaces that could be abused to cause disruption or amplify DoS.