Transit Systems Vulnerable to Disinformation

In 2023, an influencer named Kai Cenat invited his 10 million followers to New York City’s Union Square for a giveaway of PlayStation 5 consoles. (The gaming systems can retail for as much as $700.) Within two hours, more than 2,000 people flooded the 6.5-acre park. Police were unprepared for the sudden influx and chaos erupted. What started as a seemingly innocent event resulted in more than 60 arrests and several injuries, and Cenat was charged with inciting a riot. 

In today’s social media-saturated society, there is no way to predict when something or someone might “go viral.” And while that is less of a problem for trends or memes that stay online, authorities are often unprepared for when those widely shared messages impact the physical world.

“I happened to be there that day with family, when all of a sudden, hundreds of teenagers were running into Union Square,” said Jose Ramirez-Marquez, associate professor and division director of Enterprise Science and Engineering at the Stevens Institute of Technology in Hoboken, N.J.. “Cars were destroyed. Windows shattered. Some merchants of the Union Square Market were affected.”

This event is exactly what Ramirez-Marquez researches. A specialist in resilience engineering—which focuses on calculating how a system may respond to a surprise—his most recent study explores how events like the one he found himself in at Union Square that day, affect public transit. He wanted to know how vulnerable these systems are to misinformation and cyberattacks.

“It’s important to make the distinction between misinformation, which simply means someone has received incorrect information by mistake, and disinformation,” Ramirez-Marquez said. “Whereas disinformation is delivered with the intent to spread false information.”

Ramirez-Marquez used the Port Authority Trans-Hudson (PATH) system that connects New Jersey and New York as his model. He and his team analyzed the PATH system's ability to absorb and adapt to these disruptions through machine learning, specifically utilizing a k-means clustering algorithm, which sifted through a deluge of social media alerts, grouping similar reports of delays and disruptions. 

Employing advanced natural language processing techniques like BERTopic and Latent Dirichlet Allocation, the system then pinpointed those clusters driven by disinformation. Finally, to see how disinformation rippled through the system, the team also built a digital replica of the PATH network, a Monte Carlo simulation, that crunched the numbers and revealed the potential cost to commuters in both time and money.

This wasn't just about measuring the immediate impact of a single event, like a train stuck in a tunnel or a station shuttered unexpectedly. The model also simulated the domino effect, predicting how passengers would react—would they crowd platforms, seek alternate routes, or simply give up and go home?—and how those reactions would, in turn, exacerbate delays and spread disruption throughout the entire network. The findings painted a stark picture. A seemingly minor station closure could add up to a staggering 16,441 lost minutes for commuters. And the pain wasn't just measured in time; it also hit wallets, with average additional costs of $18.13 per rider as they scrambled for taxis, buses, or other alternatives. (By contrast, fares on the PATH are $3.)

The real danger emerged with higher volumes of fake news, especially if timed strategically to coincide with a mass gathering like a rally or a concert. In such a scenario, stations could be shut down as much as 11 percent of the time, unleashing a wave of delays and economic fallout across the entire region.

Beyond this specific simulation, Ramirez-Marquez emphasizes the need for city planners to consider potential cyber threats to all forms of transit, including buses and roadways. One scenario involves false alerts distributed through navigation systems, directing people onto alternate routes and creating unnecessary congestion in areas where bad actors may have other intentions.

“The way these routing algorithms work, they use real-time data,” Ramirez-Marquez said. “So, if there's a concerted effort to block some type of route then you can easily create a bottleneck somewhere else.”

Digging deeper, Ramirez-Marquez explains that these infiltrations have already happened on several occasions. Of course, this is a modern-day problem as people have become increasingly dependent on technology in the past 15 years. Rather than asking neighbors or walking up the street to see a disruption for themselves, people will look to their phones, and they cannot always fact-check the information they receive.

Looking ahead, the team aims to develop AI models capable of detecting and verifying threats across a range of systems, extending beyond transit. Our increasing dependence on technology amplifies our vulnerability to its manipulation, making Ramirez-Marquez's research all the more critical. It underscores the urgent need for a proactive approach to resilience engineering—one that anticipates both technical malfunctions and the insidious effects of disinformation.

Cassandra Kelly is a technology writer in Columbus, Ohio.