Provenance-Aware Trust Framework for Autonomous Vehicles: A Generative AI-Inspired Hybrid Approach for Decentralized Information Validation
Conference
Chowdhury, NMI, Haider, MZ, Rahman, MA et al. (2026). Provenance-Aware Trust Framework for Autonomous Vehicles: A Generative AI-Inspired Hybrid Approach for Decentralized Information Validation
. 2923-2932. 10.1109/COMPSAC69091.2026.00440
Chowdhury, NMI, Haider, MZ, Rahman, MA et al. (2026). Provenance-Aware Trust Framework for Autonomous Vehicles: A Generative AI-Inspired Hybrid Approach for Decentralized Information Validation
. 2923-2932. 10.1109/COMPSAC69091.2026.00440
Autonomous vehicles (AVs) increasingly rely on information from diverse, decentralized sources, such as peerto-peer networks, vehicle-to-vehicle communications, third-party applications, and infrastructure providers. While this information enables safer navigation, real-time decision-making, and software maintenance, it also introduces critical security challenges related to the authenticity and trustworthiness of the data received. Provenance is the ability to trace the origin and history of data, which offers a foundation for addressing these challenges. However, in highly distributed environments, cryptographic provenance alone is insufficient, as malicious or compromised entities can still propagate misleading or cryptographically valid yet falsified information. This paper proposes a provenanceaware trust framework that combines provenance verification, reputation-based evaluation, and behavioral anomaly detection with a two-stage AI-driven validation agent. The agent integrates a fine-tuned transformer-based classifier for real-time behavioral pattern recognition with a structured contextual reasoning layer for evidence fusion and trust score generation. Together, these components enable AVs to assess the trustworthiness of incoming information by jointly evaluating source provenance, behavioral consistency, and peer credibility. Unlike purely cryptographic approaches, the framework detects subtle anomalies and malicious intent even from cryptographically legitimate but compromised sources. The proposed approach aims to improve accountability and safe adoption of decentralized information, ultimately enhancing the security and trustworthiness of connected autonomous vehicles.
