Abstract
In this paper, a novel mathematical framework is introduced for modeling and analyzing the cyber-physical security of time-critical UAV applications. A general UAV security network interdiction game is formulated to model interactions between a UAV operator and an interdictor, each of which can be benign or malicious. In this game, the interdictor chooses the optimal location(s) from which to target the drone system by interdicting the potential paths of the UAVs. Meanwhile, the UAV operator responds by finding an optimal path selection policy that enables its UAVs to evade attacks and minimize their mission completion time. New notions from cumulative prospect theory (PT) are incorporated into the game to capture the operator's and the interdictor's subjective valuations of mission completion times and perceptions of the risk levels facing the UAVs. The equilibrium of the game, with and without PT, is then analytically characterized and studied, while providing detailed derivations of mission completion times (and their expected values), PT valuations of both players (along with proofs of their convergence), and equilibrium points under different studied security regimes. Novel algorithms are then proposed to reach the game's equilibria under both PT and classical game theory. Simulation results show the properties of the equilibrium for both the rational and PT cases, highlighting the effects of bounded rationality on the interdictor's and the operator's strategies as well as mission completion times, and the way it can be exploited by a fully rational opponent.
Introduction
As Unmanned Aerial Vehicles (UAVs) become central to time-critical missions—such as emergency medical deliveries or organ transport—their vulnerability to cyber-physical attacks (e.g., GPS spoofing, jamming, or physical interdiction) becomes a critical bottleneck. Traditional security models often rely on Classical Game Theory, assuming "perfectly rational" players who calculate risk using objective expected values. However, human decision-makers exhibit "bounded rationality," often overweighting rare risks and underweighting high-probability events. This research addresses the gap between idealized mathematical models and the subjective human perceptions that influence the design and oversight of autonomous systems.
System Model and Methodology
The interaction is modeled as a Network Interdiction Game with a leader-follower (Stackelberg) structure. The Interdictor (Leader) selects optimal locations in a security graph to launch attacks, aiming to maximize the UAV's mission completion time. The UAV Operator (Follower) responds with a path selection policy to minimize expected delivery time.
Key Innovations
- MDP Formulation: For mixed strategies, the operator’s problem is equivalent to finding an optimal policy in a Markov Decision Process (MDP).
- Cumulative Prospect Theory (CPT): The framework incorporates a Value Function (measuring gains/losses relative to a target time) and a Weighting Function (modeling subjective probability distortion).
- All-Paths Method: An efficient algorithm leveraging "phase-connected" security graphs to find equilibria faster than standard value iteration.
The model was validated using simulated security graphs, such as the 10-node directed graph shown above, where travel times and attack success probabilities ($p_n$) were drawn from uniform distributions to create a tractable set of examples.
Key Insights and Results
The study reveals several critical phenomena driven by human subjectivity:
- The Reference Point Effect: Strategies shift drastically based on target delivery times. Ambitious targets can drive "risk-seeking" behavior (taking dangerous shortcuts), while easily met targets lead to "risk-aversion."
- The "PT Penalty": Acting according to CPT perceptions resulted in a 14.4% increase in expected delivery time compared to a rational baseline in specific scenarios.
- Equilibrium Shift: Under CPT, interdictors move from targeting single bottlenecks to more distributed strategies, anticipating the operator's subjective fears.
- Exploiting Irrationality: A fully rational operator can outperform a PT-operator by over 30% by anticipating the attacker's subjective biases.
Conclusion
This framework demonstrates that subjective perceptions of risk and time-delays significantly alter strategic outcomes in UAV security. While bounded rationality generally increases mission completion times, understanding these biases allows for more robust system design. Future work will extend this model to dynamic graphs and multi-player scenarios involving drone swarms.
Summary updated