Explainable and trustworthy AI-driven fuzzy consensus for secure data sharing in intelligent transportation systems

In the bustling arteries of a modern city, cars are no longer just machines; they are moving computers constantly exchanging data to avoid accidents and ease traffic. This network of connected vehicles, often called the Internet of Vehicles, relies on a simple but fragile premise: that the information a car shares is true and that the car sharing it can be trusted. When a vehicle suddenly brakes or reports a hazard, the cars behind it must decide instantly whether to believe that signal. If they trust a liar, the result can be a cascade of confusion or a crash. The challenge for engineers is to build a system that verifies this information quickly enough to be useful, without exposing the private details of every driver's route or identity. To solve this, researchers are turning to a blend of artificial intelligence and mathematical logic that mimics how humans handle uncertainty, moving away from rigid, black-box calculations toward a system that can explain its own reasoning.

A team of researchers from Zhejiang University of Technology and China Overseas Property Management has proposed a new way to handle this trust problem. They argue that traditional methods for checking if a vehicle is reliable are often too rigid or too slow for the chaotic reality of city traffic. Instead of relying on a simple score that says a car is "good" or "bad," their approach uses a concept called intuitionistic fuzzy logic. This method allows the system to acknowledge that trust is not a fixed number but a shifting state that includes doubt and hesitation. By translating complex driving histories and real-time locations into human-understandable language like "high trust" or "low trust," the system can make decisions that are both fast and transparent. The researchers built a framework where vehicles and roadside units work together to verify messages, using this fuzzy logic to select the most reliable participants for the job while keeping sensitive data hidden.

The core of their work is a three-stage process that begins when a vehicle enters a network. First, the system looks at the car's past behavior, not by crunching raw numbers, but by converting its history of successful data sharing or rule-breaking into a fuzzy trust profile. If a car has a history of reporting accidents correctly, it gains a high degree of membership in the "trustworthy" category. If it has been caught sending false data, its trust level drops. Crucially, this process does not require the car to reveal its entire history in a way that could be stolen; the data is abstracted into semantic categories that are safe to share. Next, the system considers the immediate context. It asks how close the car is to the event and whether its path aligns with the traffic flow. A car that is physically near a traffic jam and moving in the same direction as the flow is given more weight than one that is far away or moving in the opposite direction. This spatial similarity is also translated into fuzzy terms like "near" or "far," ensuring that the system understands the situation without needing to know the exact GPS coordinates of every driver.

Finally, these two pieces of information—the car's history and its current situation—are combined to make a decision. The system selects a small group of vehicles to vote on whether a message is true. Unlike older systems where every vote might count the same, or where a few powerful nodes decide everything, this new method weights the votes based on the fuzzy trust scores. A vehicle with a high trust score and a relevant location has more influence on the final decision. If the group agrees that the message is valid, it is recorded on a secure, unchangeable digital ledger, creating a permanent record of the event. If the message is flagged as suspicious, it is isolated to prevent it from spreading. The researchers tested this system in a simulated environment with one hundred vehicles and ten roadside units, mimicking the density and speed of a real city. They found that their method could complete one hundred rounds of verification in just twelve seconds, significantly faster than other leading methods which took nearly twenty-nine to forty-eight seconds for the same task.

Beyond speed, the study demonstrated that this approach is much better at spotting bad actors. In simulations where thirty percent of the vehicles were programmed to act maliciously by sending false data, the new system successfully identified and filtered out the bad information more often than systems that relied only on past behavior or simple statistical models. The fuzzy logic allowed the system to adapt to the chaos of the moment, recognizing that a car with a good history might still be unreliable if it was far away from the event, or that a car with a questionable past might be telling the truth if it was right at the scene. Furthermore, the system offered a stronger shield for privacy. When researchers tried to use prediction models to reconstruct the exact paths of the vehicles from the data the system shared, they found that the fuzzy representation made it much harder to guess where a car had been. The error in predicting the car's location was significantly higher than with raw data, meaning that drivers could share necessary traffic information without revealing their precise movements.

The researchers emphasize that this is not a magic solution that eliminates all risk, but a robust framework designed to make intelligent transportation systems more accountable and secure. By replacing opaque algorithms with a system that can explain its decisions in plain terms, the work offers a path toward a future where cars can coordinate safely without sacrificing the privacy of their passengers. The simulations suggest that this balance of speed, security, and transparency is achievable, providing a foundation for the next generation of smart cities where data sharing is not just fast, but also trustworthy.