Progress Update
The communication attack suite was expanded beyond basic false-data injection. The eventual evaluated set included clean communication, FDI-position, FDI-velocity, FDI-acceleration, replay, fixed/random delay, packet loss, collusion, stealth perturbations, and mixed attacks.
Experiment Note
Each attack was designed to stress a distinct failure mode:
- FDI attacks directly falsified communicated kinematic fields.
- Replay reused previously plausible states, making simple range checks insufficient.
- Delay made messages temporally stale while leaving their values internally plausible.
- Packet loss removed information rather than corrupting it.
- Collusion corrupted multiple sources in a coordinated way.
- Stealth attacks kept perturbations small enough to challenge simple detectors.
- Mixed attacks tested whether the architecture generalized beyond one isolated corruption mechanism.
Technical Essay
This distinction later became important in interpreting results. Trust weighting is naturally effective when information is present but unreliable. It cannot reconstruct information that is absent. Packet loss therefore exposed a fundamental limitation rather than merely a tuning problem. Delay created a related failure: a state can be physically plausible and still be useless because it is stale.
The threat model assumed communication compromise but not direct actuator compromise or compromised local range sensing. This preserved a trustworthy physical channel for the runtime shield. If both V2V and onboard physical sensing were compromised, the proposed shield would no longer have the information required to enforce gap/TTC constraints.