Don't Let One Lie Survive A Hundred Truths: A Selective Bayesian Trust Estimator for Collaborative Perception
SABER detects stealthy single-object attacks in vehicle collaborative perception while lowering benign false positives.
SABER is a selective two-tier Bayesian trust estimator for collaborative perception in connected vehicles, aimed at collaborators that hide or fabricate a single object while reporting others correctly. Cumulative-sum screening selects persistent agent-object anomalies for a separate reference-weighted Beta assessment, and the lowest pair score constrains agent trust. On OPV2V, defense ROC-AUC improves over MATE by up to 0.427 in late fusion and 0.337 in intermediate fusion. Against advanced intermediate-fusion fabrication, detection rises by up to 96.40 and 67.07 percentage points over ROBOSAC and LUCIA while benign false-positive rates drop.
- Targets stealthy single-object omission or injection by collaborators.
- A separate Beta state prevents unrelated correct reports from diluting evidence.
- OPV2V ROC-AUC rises by up to 0.427 over MATE in late fusion.
- Detection gains reach 96.40 points over ROBOSAC on fabrication attacks.
- Benign false-positive rates fall versus prior collaborative-perception defenses.
From the original article
1 min read at arxiv.orgCollaborative perception (CP) enables connected vehicles to see beyond their own sensors but makes them dependent on messages they cannot independently verify. A compromised collaborator can surgically conceal a single safety-critical object or inject a non-existing one while correctly reporting many others.
By Yutong Liu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.