Website: https://ans.unibs.it/docs/plexeMDS.html
The ANS lab developed a Hybrid Misbehavior Detection System (MDS) for Plexe, the platooning extension of the SUMO/VEINS vehicular network simulator. The methodology is described in the paper: L. Ghiro, C. Pezzoni, M. Franceschini, R. Lo Cigno — “Physics Joins AI: A Real-Time Hybrid Misbehavior Detection Framework for Vehicular Networks” [Preprint].
 |
|---|
| The same simulation of a platoon under attack: (above) without an MDS and (below) with an MDS. Without the MDS the platoons under attack crashes while, enabling the MDS, the platoons detects the attacks and reacts by opening safety gaps. |
Vehicular platooning promises significant safety and efficiency benefits, but it also introduces new attack surfaces: a single misbehaving vehicle injecting false data into V2X beacons can cause catastrophic chain collisions across an entire platoon. Existing detection approaches rely either on AI models — powerful but opaque and dependent on representative training data — or on physics-based rules — interpretable but brittle against novel attacks.
PlexeMDS combines these two complementary strategies into a single real-time framework. An LSTM-based neural network analyses sequences of received beacons to spot statistical anomalies, while a parallel physics coherence check verifies that the reported kinematic data are physically plausible. The two verdicts are fused by a lightweight decider that allows either method to trigger an alarm, maximising recall while keeping false-positive rates low.
The system is evaluated both offline on the VeReMi dataset, covering a wide range of known attack types, and in runtime highway platooning simulations inside Plexe, measuring collision rate reduction and reaction time. The results show that enabling the Hybrid MDS is sufficient to prevent crashes that would otherwise occur, while the platoon seamlessly reacts by opening safety gaps upon attack detection.
The source code and full documentation are available on GitHub:
- plexe/comcomvsi2025-plexe3.2 - PLEXE, featuring the Hybrid MDS implementation.
- veins/comcomvsi2025 - VEINS, updated for compatibility with TraCI 22;
- sumo/comcomvsi2025-plexe3.2 - SUMO, extended to support front-gap computation using radar or beacon data for CACC algorithms;