Reinforcement learning (RL) has demonstrated remarkable success in solving complex decision-making problems, yet its adoption in critical domains is hindered by the lack of interpretability in its decision-making processes. Existing explainable AI (xAI) approaches often fail to provide meaningful explanations for RL agents, particularly because they overlook the contrastive nature of human reasoning—answering "why this action instead of that one?" To address this gap, we propose a novel framework of contrastive learning to explain RL selected actions, named VisionMask. VisionMask is trained to generate explanations by explicitly contrasting the agent's chosen action with alternative actions in a given state using a self-supervised manner. We demonstrate the efficacy of our method through experiments across diverse RL environments, evaluating it in terms of faithfulness, robustness and complexity. Our results show that VisionMask significantly improves human understanding of agent behavior while maintaining accuracy and fidelity. Furthermore, we present examples illustrating how VisionMask can be used for counterfactual analysis. This work bridges the gap between RL and xAI, paving the way for safer and more interpretable RL systems.

In Cyber-Physical Systems (CPS), sensor data integrity is crucial since acting on malicious sensor data can cause serious consequences, given the tight coupling between cyber components and physical systems. While extensive works focus on sensor attack detection, attack diagnosis that aims to find out when the attack starts has not been well studied yet. This temporal sensor attack diagnosis problem is equally important because many recovery methods rely on the accurate determination of trustworthy historical data. To address this problem, we propose a lightweight data-driven solution to achieve real-time sensor attack diagnosis. Our novel solution consists of five modules, with the attention and diagnosis ones as the core. The attention module not only helps accurately predict future sensor measurements but also computes statistical attention scores for the diagnosis module. Based on our unique observation that the score fluctuates sharply once an attack launches, the diagnosis module determines the onset of an attack through monitoring the fluctuation. Evaluated on high-dimensional high-fidelity simulators and a testbed, our solution demonstrates robust and accurate temporal diagnosis results while incurring millisecond-level computational overhead on Raspberry Pi.

Real-time systems are susceptible to adversarial factors such as faults and attacks, leading to severe consequences. This paper presents an optimal checkpoint scheme to bolster fault resilience in real-time systems, addressing both logical consistency and timing correctness. First, we partition message-passing processes into a directed acyclic graph (DAG) based on their dependencies, ensuring checkpoint logical consistency. Then, we identify the DAG’s critical path, representing the longest sequential path, and analyze the optimal checkpoint strategy along this path to minimize overall execution time, including checkpointing overhead. Upon fault detection, the system rolls back to the nearest valid checkpoints for recovery. Our algorithm derives the optimal checkpoint count and intervals, and we evaluate its performance through extensive simulations and a case study. Results show a 99.97% and 67.86% reduction in execution time compared to checkpoint-free systems in simulations and the case study, respectively. Moreover, our proposed strategy outperforms prior work and baseline methods, increasing deadline achievement rates by 31.41% and 2.92% for small-scale tasks and 78.53% and 4.15% for large-scale tasks.

Sensor attacks alter sensor readings and spoof Cyber-Physical Systems (CPS) to perform dangerous actions. Existing detection works tend to minimize the detection delay and false alarms at the same time, while there is a clear trade-off between the two metrics. Instead, we argue that attack detection should dynamically balance the two metrics when a physical system is at different states. Along with this argument, we propose an adaptive sensor attack detection system that consists of three components - an adaptive detector, detection deadline estimator, and data logger. It can adapt the detection delay and thus false alarms at run time to meet a varying detection deadline and improve usability (or false alarms). Finally, we implement our detection system and validate it using multiple CPS simulators and a reduced-scale autonomous vehicle testbed.

This paper proposes an optimal checkpoint scheme for fault resilience in real-time systems, in which we consider both logical consistency and timing correctness. First, we partition message-passing processes into a directed acyclic graph (DAG) considering their dependencies, where the logical consistency of checkpoints is guaranteed. Then, we find the critical path of the DAG, which is the longest path performed in sequence. Next, we analyze the optimal checkpoint strategy on the critical path where the overall execution time (including checkpointing overhead) is minimized. When a fault is detected, the system rolls back to the nearest valid checkpoint for recovery. The optimal number of checkpoints and their intervals are derived by the algorithm.