LAUSR

Peer‐to‐peer semantic vehicular networks (P2P‐SVNs) use semantic communication to enable efficient vehicle‐to‐vehicle interaction by exchanging meaning instead of raw data. However, P2P‐SVNs face unique backdoor threats where attackers embed triggers in semantic features. Current research predominantly focuses on explicit trigger patterns that can be identified and filtered by vehicle IDS. To address this shortcoming, we propose IBPS, a novel backdoor attack framework for P2P‐SVN systems that utilises subliminal triggers to evade detection. The attack operates through two key phases: (1) training phase where latent backdoors are embedded into the semantic encoder via adversarial trigger poisoning and (2) inference phase where these backdoors are automatically activated by subconscious trigger while maintaining normal communication performance. Experimental validation on P2P‐SVNs and benchmark datasets confirms that IBPS effectively maintains high attack performance while preserving normal functionality and exhibiting strong stealth.