Abstract
Vehicular Ad-Hoc Networks (VANETs) are poised to enhance road safety through real-time data exchange. However, their decentralized nature makes them vulnerable to malicious misinformation and network congestion. This paper presents a decentralized trust-based framework that integrates local sensor verification with a weighted majority voting system. By employing cluster-based routing and a hybrid trust model—combining static role-based trust with dynamic experience-based metrics—the proposed system effectively filters malicious messages and reduces spam. Simulation results demonstrate a 65% decrease in erroneous driver actions and restricted spam propagation, ensuring both network scalability and information integrity.
Introduction
As vehicles transition into nodes within massive, high-speed networks, the promise of Vehicular Ad-Hoc Networks (VANETs) to revolutionize road safety becomes tangible. By swapping data regarding accidents and hazards, cars can proactively manage traffic and prevent collisions. Yet, the lack of a central authority in these dynamic environments introduces significant risks: malicious actors or faulty sensors can broadcast false safety alerts, leading to traffic manipulation or accidents. Furthermore, unrestricted broadcasting (flooding) can saturate limited wireless bandwidth with "spam," blocking critical safety data. This research addresses these challenges by developing a "firewall" of trust that operates at both the network relaying level and the individual decision-making level.
Methodology
The proposed framework is built upon three fundamental pillars designed to ensure scalability and reliability:
1. Cluster-Based Propagation
To maintain network scalability, vehicles are organized into geographical clusters. Rather than universal broadcasting, "cluster leaders" aggregate trust opinions and determine if a message warrants relaying to the next hop. This hierarchical structure prevents the network from being overwhelmed by redundant or untrusted data.
2. Hybrid Trust Model
The system evaluates peer reliability using two distinct metrics:
- Role-based Trust: Static, high trust values are assigned to recognized authorities, such as police vehicles or road-side units.
- Experience-based Trust: For standard vehicles, trust is dynamic and calculated using a recursive formula with a "forgetting factor." This allows the system to reward consistent accuracy and penalize misleading behavior over time.
3. Analysis and Action Modules
Each vehicle is equipped with an Analysis Module to compare incoming data against local sensor readings and an Action Module to aggregate weighted opinions from peers. To optimize bandwidth, the framework utilizes identity-based aggregate signatures, compressing multiple trust opinions into a single verifiable package.
Experimental Setup
The framework was evaluated using a custom C++ simulation environment modeling the East York area of Toronto. The simulation included 1,125 vehicle entities following realistic traffic dynamics and road limits. Two primary threat models were tested: "spammers" broadcasting high-frequency junk and "malicious peers" sending deceptive safety information.
Results and Key Insights
The simulation compared the proposed framework against a baseline of unrestricted relaying and partial trust models.
Spam Control and Scalability
Trust is essential for network health. By filtering untrusted messages at the cluster leader level, the system prevents malicious data from clogging communication channels. Without relay control, spam could propagate up to 5.5 km. The proposed framework restricted this to approximately 2 km using role-based trust alone, with further reductions as experience-based trust matured.
Decision Accuracy and System Recovery
The integration of trust opinions led to a 65% reduction in "wrong actions"—instances where a vehicle follows a false message. A key finding was the resilience of experience-based trust; even in scenarios with 10% malicious peers, the network's relay effectiveness began to recover after approximately 35 minutes as the system identified and sidelined bad actors.
Conclusion
This trust-based framework provides a robust solution for securing VANETs against misinformation and congestion. While effective, future work is needed to address the "malicious leader" problem through revocation mechanisms and to adapt the system for extreme traffic densities. Additionally, the real-world implementation of the Analysis Module remains a critical engineering hurdle for future smart road infrastructure.
Summary updated