Traffic networks are vulnerable to disinformation attacks

Marcin Waniek, Gururaghav Raman, Bedoor K. AlShebli, Jimmy Chih-Hsien Peng, Talal RahwanarXiv 2020 · arXiv · OpenReview

TL;DR

This paper explores a novel attack vector on critical infrastructure: using disinformation to manipulate driver behavior and cause traffic congestion. They demonstrate that even a small percentage of drivers following fake traffic alerts can lead to significant disruptions, and surprisingly, concentrating attacks in a business district can have a wider reach than spreading them across a city.

What Problem Does This Paper Address?

The increasing sophistication and pervasiveness of disinformation campaigns pose a growing threat to society. While much attention has been given to their impact on public opinion and political discourse, their potential to disrupt critical infrastructure has been largely overlooked. This paper specifically investigates the vulnerability of urban traffic networks to disinformation-based attacks. The core problem is understanding how fake information, designed to influence individual drivers' route choices, can be weaponized to create widespread traffic congestion and disruption, and to what extent this manipulation can be effective.

Methods

The authors employ a multi-pronged approach:

  • Survey: They conducted a survey on Amazon Mechanical Turk with over 3,300 participants to gauge their propensity to follow and forward fake traffic alerts.
  • Traffic Simulation: They developed a sophisticated cellular-automata traffic model (a modified Nagel-Schreckenberg model) to simulate traffic flow in Chicago. This model accounts for individual vehicle behavior, road networks, and the impact of rerouting.
  • Computational Complexity Analysis: They formulated the adversary's problem as "Maximizing Disruption" and "Minimizing Targets," and analyzed their computational complexity, proving them to be NP-complete and APX-hard, respectively. This implies that finding an optimal attack strategy is computationally intractable.
  • Heuristic Analysis: Given the intractability of optimal solutions, they analyzed a greedy heuristic for selecting attack targets and simulated its impact under various disinformation follow-through rates.

Datasets

  • OpenStreetMap (OSM) data: Used to generate the road network of Chicago.
  • City of Chicago traffic data: Provided daily average vehicle counts for locations.
  • Texas A&M Transportation Institute data: Used for the distribution of traffic intensity throughout the day.
  • Survey Data: Collected from 3,301 participants on their likelihood to follow and forward fake traffic alerts.

Key Insights and Ideas

  • Behavioral Vulnerability: Critical infrastructure, like traffic networks, can be vulnerable not just to hardware/software exploits, but also to the manipulation of human behavior through disinformation.
  • High Receptivity to Fake Alerts: A significant majority (89%) of survey participants indicated a high propensity to follow fake traffic alerts, suggesting a widespread susceptibility to such manipulations.
  • Concentrated Attacks Amplify Impact: Counter-intuitively, concentrating disinformation attacks on a specific area (like a business district) can lead to more far-reaching disruption than spreading them across the entire city.
  • Computational Intractability of Optimal Attacks: Finding the "perfect" set of roads to target for maximum disruption is computationally infeasible, forcing adversaries to rely on heuristics.
  • Braess' Paradox in Action: The simulations showed instances of Braess' paradox, where rerouting due to an attack could paradoxically decrease travel times for some drivers while increasing it for others.

Key Results

  • Significant Disruption Potential: Even with a 50% follow-through rate, disinformation attacks can cause substantial changes in traffic flow, with impacts felt up to 2 kilometers away from targeted areas.
  • Concentrated Attack Reach: When targets were confined to Chicago's Loop (business district), the disruption's impact was felt up to 2 kilometers away, a wider reach than when targets were spread across the city.
  • Increased Variance in Delays: Higher follow-through rates lead to a greater variance in ride times, meaning more drivers experience significantly longer delays. For instance, a 75% follow-through rate increased the number of rides with delays of 60+ minutes from ~1,100 to ~29,300.
  • Heuristic vs. Random Attacks: The greedy heuristic for selecting targets resulted in a more skewed distribution of delays compared to random attacks, highlighting the importance of the adversary's strategy.

Limitations and Future Work

  • Survey Realism: The authors acknowledge that survey responses might not perfectly reflect real-world behavior.
  • Shortest Path Assumption: The simulation used shortest paths for rerouting, whereas real drivers might prioritize fastest paths, which would require more complex dynamic recalculations.
  • Ethical Concerns of Field Experiments: While a field experiment could offer more realistic data, ethical and safety concerns related to manipulating unwitting drivers are significant.
  • Future Work: The authors suggest extending the functionality of navigation apps to include crowdsourced verification of traffic incidents. They also highlight the need for effective detection and counter-measures against such disinformation-based attacks.

Summary updated

Abstract

Disinformation continues to attract attention due to its increasing threat to society. Nevertheless, a disinformation-based attack on critical infrastructure has never been studied to date. Here, we consider traffic networks and focus on fake information that manipulates drivers' decisions to create congestion. We study the optimization problem faced by the adversary when choosing which streets to target to maximize disruption. We prove that finding an optimal solution is computationally intractable, implying that the adversary has no choice but to settle for suboptimal heuristics. We analyze one such heuristic, and compare the cases when targets are spread across the city of Chicago vs. concentrated in its business district. Surprisingly, the latter results in more far-reaching disruption, with its impact felt as far as 2 kilometers from the closest target. Our findings demonstrate that vulnerabilities in critical infrastructure may arise not only from hardware and software, but also from behavioral manipulation.
BibTeX
@misc{waniek2020traffic,
  author = {Marcin Waniek and Gururaghav Raman and Bedoor K. AlShebli and Jimmy Chih-Hsien Peng and Talal Rahwan},
  title = {{Traffic networks are vulnerable to disinformation attacks}},
  year = {2020},
  eprint = {2003.03723},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2003.03723},
}

Loading related content...

Figure