Safe Reinforcement Learning (RL) has been applied to synthesize control policies that maximize task rewards while adhering to safety constraints within simulated secure cyber-physical systems. However, the vulnerability of safe RL to adversarial attacks remains largely unexplored. We argue that understanding the safety vulnerabilities of learned control policies is crucial for ensuring true safety in real-world scenarios. To address this gap, we first formally define the safe RL problem with formal language (signal temporal logic) and demonstrate that even optimal policies are susceptible to observation perturbations. We then introduce novel safety violation attacks that exploit adversarial models trained with reversed safety constraints to induce unsafe behaviors. Lastly, through both theoretical analysis and experimental results, we demonstrate that our approach is more effective at violating safety constraints than existing adversarial RL methods, which primarily focus on reducing task rewards rather than compromising safety.

We study the vulnerability of Cyber-physical systems (CPS) under stealthy sensor attacks in black-box scenarios. “Black-box” refers to scenarios where the attacker has minimal knowledge of the target system. Designing a stealthy sensor attack sequence under this scenario has two main challenges. The first one lies in ensuring the stealthiness of the sensor attack, meaning does not trigger an alert when applying the generated sensor attack sequence to the CPS. The second one is maintaining stealthiness throughout the attack generation process, indicating the limitation on the alarm frequency when generating the attack sequence. To address the above challenges, we develop a querybased black-box stealthy attack framework to violate the safety of the CPS. To maintain stealthiness during training, an active learning method has been introduced to extract the detector’s information to a time series model. The stealthy attack sequence is then generated from that model. Experiments on four numerical simulations and a high-fidelity simulator demonstrate the effectiveness of the proposed framework.

Sensor attacks on Cyber-Physical Systems (CPS) can cause substantial damage in the physical world, which motivates two major threads of defense works including attack detection and attack recovery. The former aims to identify whether any sensors are compromised while the latter seeks to restore a system to safety once an attack is detected. Although either thread has drawn many efforts, how to coordinate the detection and recovery has been barely studied. Overlooking the coordination, existing works may result in ineffective and even failed defense. For example, if a detector raises an alarm too late, there may not be enough time for a system to recover but reach the unsafe region anyway, even though the detection result is accurate. By contrast, raising an alarm earlier allows more time for recovery, but may come with more false positives and thus unnecessarily trigger the recovery. To fill this gap, we aim to co-design attack detection and recovery, and propose a novel recovery-guaranteed sensor attack detection framework. The framework dynamically adjusts the detection sensitivity and authenticates state estimates at run time to guarantee timely and safe recovery once an attack is detected. The detection will always reserve sufficient time for the recovery while minimizing unnecessary activation of recovery. We conduct extensive simulations and real-world testbed experiments to show the efficiency of our solution.

Reaching a target safely and quickly is a control goal pursued by various applications, such as post-disaster rescue robots and industrial shipment. However, it is hard to formally guarantee safety and time-optimality under unknown dynamics via model-free controller synthesis algorithms. As a response, we propose a model-free reinforcement learning (RL) algorithm that synthesize a controller to reach a predefined target set of states with a probabilistic guarantee of time optimality, i.e., the actual reaching time is bounded close to the shortest time possible with high probability, and the bound becomes tighter when more training data is sampled. Our algorithm leverages a reward function that based on signal temporal logic (STL) robustness to reward fast reaching. With this reward function, we prove that Probably Approximately Correct (PAC) optimality in the state-value function implies PAC optimality in reach time. Then, we build our algorithm by extending Deplayed Gaussian Process Q learning (DGPQ) algorithm with a safety margin to protect the controlled agent. Consequently, our algorithm guarantees safety and a PAC bound in recovery time. Experiments show our method can achieve $\mathbf{9 7. 7 \%}$ success rate to reach the target with in the maximum time tolerance and outperform baselines.

Cyber-Physical Systems(CPS) are the integration of sensing, control, computation, and networking with physical components and infrastructure connected by the internet. The autonomy and reliability are enhanced by the recent development of safe reinforcement learning (safe RL). However, the vulnerability of safe RL to adversarial conditions has received minimal exploration. In order to truly ensure safety in physical world applications, it is crucial to understand and address these potential safety weaknesses in learned control policies. In this work, we demonstrate a novel attack to violate safety that induces unsafe behaviors by adversarial models trained using reversed safety constraints. The experiment results show that the proposed method is more effective than existing works.

Industries are embracing information technology and constructing more robust machines known as Cyber-Physical Systems(CPS) to automate processes. CPSs are envisioned to be pervasive, coordinating, and integrating computation, sensing, actuation, and physical processes. CPSs have various applications in life-critical scenarios, where their performance and reliability can have direct impacts on human safety and well-being. However, CPSs are vulnerable to malicious attacks, and researchers have developed detectors to identify such attacks in different contexts. Surprisingly, little work has been done to detect attacks on the actuators of CPS. Furthermore, actuators face a high risk of optimal hidden attacks designed by powerful attackers, which can push them into an unsafe state without detection. To the best of our knowledge, no such attacks on actuators have been developed yet. In this paper, we design an optimal hidden attack for actuators and evaluate its effectiveness. First, we develop a mathematical model for actuators and then create a linear program for convex optimization. Second, we solve the optimization problem and simulate the optimal attack.

Cyber-physical systems tightly integrate computational resources with physical processes through sensing and actuating, widely penetrating various safety-critical domains, such as autonomous driving, medical monitoring, and industrial control. Unfortunately, they are susceptible to assorted attacks that can result in injuries or physical damage soon after the system is compromised. Consequently, we require mechanisms that swiftly recover their physical states, redirecting a compromised system to desired states to mitigate hazardous situations that can result from attacks. However, existing recovery studies have overlooked stochastic uncertainties that can be unbounded, making a recovery infeasible or invalidating safety and real-time guarantees. This paper presents a novel recovery approach that achieves the highest probability of steering the physical states of systems with stochastic uncertainties to a target set rapidly or within a given time. Further, we prove that our method is sound, complete, fast, and has low computational complexity if the target set can be expressed as a strip. Finally, we demonstrate the practicality of our solution through the implementation in multiple use cases encompassing both linear and nonlinear dynamics, including robotic vehicles, drones, and vehicles in high-fidelity simulators.

Meeting deadlines is a fundamental requirement of cyber-physical systems (CPS) in real-time applications to consolidate their reliability and effectiveness in executing timecritical tasks. Recent research have focused on applying reinforcement learning to synthesize controllers for real-time systems, particularly in terms of achieving fast reach-avoid. However, achieving fast behavior does not necessarily equate to meeting deadlines. Sometimes reinforcement learning agents are trying to maximize the total reward by exploiting the reward function, and thus performing unwanted behavior, known as reward hacking. Therefore, depending on the deadlines, it is possible to have fast controllers that miss the deadlines and slow controllers that meet the deadlines. To address the misalignment between fast and meeting deadlines, we investigate the relationship between as soon as possible (ASAP) and deadline-safe. Additionally, we formulate the problem into a new Markov decision process R-MDP including time to avoid non-Markovian rewards when considering deadlines. Furthermore, we have designed new reward functions that encourage the agent to meet the deadlines. Moreover, we evaluate our method on various benchmarks. The experiment results show the effectiveness of our method in ensuring deadline compliance without compromising safety.

Safe reinforcement learning (RL) aims to derive a control policy that navigates a safety-critical system while avoiding unsafe explorations and adhering to safety constraints. While safe RL has been extensively studied, its vulnerabilities during the policy training have barely been explored in an adversarial setting. This article bridges this gap and investigates the training time vulnerability of formal language-guided safe RL. Such vulnerability allows a malicious adversary to inject backdoor behavior into the learned control policy. First, we formally define backdoor attacks for safe RL and divide them into active and passive ones depending on whether to manipulate the observation. Second, we propose two novel algorithms to synthesize the two kinds of attacks, respectively. Both algorithms generate backdoor behaviors that may go unnoticed after deployment but can be triggered when specific states are reached, leading to safety violations. Finally, we conduct both theoretical analysis and extensive experiments to show the effectiveness and stealthiness of our methods.

Unmanned aerial vehicles (UAVs) have various applications in different settings, including for example, surveillance, packet delivery, emergency response, data collection in the Internet of Things (IoT), and connectivity in cellular networks. However, this technology comes with many risks and challenges such as vulnerabilities to malicious cyber-physical attacks. This article studies the problem of path planning for UAVs under GPS sensor permanent faults in a cyber-physical system (CPS) perspective. Based on studying and analyzing the CPS architecture of the UAV, the cyber “attacks and threats” are differentiated from attacks on sensors and communication components. An efficient way to address this problem is to introduce a novel approach for UAV’s path planning resilience to cyber-attack artificial potential field (RCA-APF) algorithm. The proposed algorithm completes the three stages in a coordinated manner. In the first stage, the permanent faults on the GPS sensor of the UAV are detected, and the UAV starts to divert from its initial path planning. In the second stage, we estimated the location of the UAV under GPS permanent fault using received signal strength (RSS) trilateration localization approach. In the final stage of the algorithm, we implemented the path planning of the UAV using an open source UAV simulator. Experimental and simulation results demonstrate the performance of the algorithm and its effectiveness, resulting in efficient path planning for the UAV.

There are various applications of Cyber-Physical systems (CPSs) that are life-critical where failure or malfunction can result in significant harm to human life, the environment, or substantial economic loss. Therefore, it is important to ensure their reliability, security, and robustness to the attacks. However, there is no widely used toolbox to simulate CPS and target security problems, especially the simulation of sensor attacks and defense strategies against them. In this work, we introduce our toolbox CPSim, a user-friendly simulation toolbox for security problems in CPS. CPSim aims to simulate common sensor attacks and countermeasures to these sensor attacks. We have implemented bias attacks, delay attacks, and replay attacks. Additionally, we have implemented various recovery-based methods against sensor attacks. The sensor attacks and recovery methods configurations can be customized with the given APIs. CPSim has built-in numerical simulators and various implemented benchmarks. Moreover, CPSim is compatible with other external simulators and can be deployed on a real testbed for control purposes. 1

Cyber-physical systems (CPS) are susceptible to physical attacks, and researchers are exploring ways to detect them. One method involves monitoring the system for a set duration, known as the time-window, and identifying residual errors that exceed a predetermined threshold. However, this approach means that any sensor attack alert can only be triggered after the time-window has elapsed. The length of the time-window affects the detection delay and the likelihood of false alarms, with a shorter time-window leading to quicker detection but a higher false positive rate, and a longer time-window resulting in slower detection but a lower false positive rate.While researchers aim to choose a fixed time-window that balances a low false positive rate and short detection delay, this goal is difficult to attain due to a trade-off between the two. An alternative solution proposed in this paper is to have a variable time-window that can adapt based on the current state of the CPS. For instance, if the CPS is heading towards an unsafe state, it is more crucial to reduce the detection delay (by decreasing the time-window) rather than reducing the false alarm rate, and vice versa. The paper presents a sensor attack detection framework that dynamically adjusts the time-window, enabling attack alerts to be triggered before the system enters dangerous regions, ensuring timely detection. This framework consists of three components: attack detector, state predictor, and window adaptor. We have evaluated our work using real-world data, and the results demonstrate that our solution improves the usability and timeliness of time-window-based attack detectors.

The paper describes the design of a simulation and security toolbox for cyber-physical systems, and demonstrates two real-time recovery cases based on the toolbox.

Cyber-physical systems (CPSs) leverage computations to operate physical objects in real-world environments, and increasingly more CPS-based applications have been designed for life-critical applications. Therefore, any vulnerability in such a system can lead to severe consequences if exploited by adversaries. In this paper, we present a data predictive recovery system to safeguard the CPS from sensor attacks, assuming that we can identify compromised sensors from data. Our recovery system guarantees that the CPS will never encounter unsafe states and will smoothly recover to a target set within a conservative deadline. It also guarantees that the CPS will remain within the target set for a specified period. Major highlights of our paper include (i) the recovery procedure works on nonlinear systems, (ii) the method leverages uncorrupted sensors to relieve uncertainty accumulation, and (iii) an extensive set of experiments on various nonlinear benchmarks that demonstrate our framework’s performance and efficiency.

While many research efforts on Cyber-Physical System (CPS) security are devoted to attack detection, how to respond to the detected attacks receives little attention. Attack response is essential since serious consequences can be caused if CPS continues to act on the compromised data by the attacks. In this work, we aim at the response to sensor attacks and adapt machine learning techniques to recover CPSs from such attacks. There are, however, several major challenges. i) Cumulative error. Recovery needs to estimate the current state of a physical system (e.g., the speed of a vehicle) in order to know if the system has been driven to a certain state. However, the estimation error accumulates over time in presence of compromised sensors. ii) Timely response. A fast response is needed since slow recovery not only comes with large estimation errors but also may be too late to avoid irreparable consequences. To address these challenges, we propose a novel learning-based solution, named sequence-predictive recovery (or SeqRec). To reduce the estimation error, SeqRec designs the first sequence-to-sequence (Seq2Seq) model to uncover the temporal and spatial dependencies among sensors and control demands, and then uses the model to estimate system states using the trustworthy data logged in history. To achieve an adequate and fast recovery, SeqRec designs the second Seq2Seq model that considers both the current time step using the remaining intact sensors and the future time steps based on a given target state, and embeds the model into a novel recovery control algorithm to drive a physical system back to that state. Experimental results demonstrate that SeqRec can effectively and efficiently recover CPSs from sensor attacks.

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.

In Cyber-Physical Systems (CPS), integrating new technologies that interact with and control physical systems raises new security risks beyond the classical cyber security domain. These risks motivated many attack detectors that focus on the binary outcome. However, one pressing risk in CPS is hidden sensor attacks that are well-designed by powerful attackers who gained full knowledge of our systems and detector. The hidden attacks inject such a small malicious signal into sensor measurement that they can stay undetected but eventually lead to a significant deviation. Thus, to secure the CPS, we propose a detection framework to identify these sensor attacks that can drive the system's physical states to an unsafe state within a given period, even if they are not detected. First, we solve optimization problems to find the optimal hidden sensor attack that leads to the minimal distance to a pre-defined unsafe state region within an observation window for a given system and detector. Then, based on this algorithm, we perform offline profiling to search for a conditionally safe region, where the system states are guaranteed to be safe within the observation window as long as the detector does not raise any alerts. Finally, the framework can online discover potential hidden sensor attacks that endanger the system by checking if the current system state moves out of the region and raising a yellow alert. The evaluation shows that the optimal hidden sensor attack results in the minimum distance to unsafe, within a given observation window among existing hidden sensor attacks. We implemented our method on four linear simulators to show the effectiveness of our method. Additionally, we provided a discussion on the challenges of applying the proposed method to non-linear systems.

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.

Malicious attackers may disrupt the safety of autonomous systems through compromising sensors to feed wrong measurements to the controller. This article proposes attack-resilient sensor fusion that combines local sensor readings and shared sensing information from multiple sources. The method results in higher resilience against sensor attacks through jointly considering sensing noise and uncertain communication delay. To be specific, we first identify the considerable impact of the delay on determining attacked sensors. Second, we present a novel two-dimensional abstract sensor model, where each measurement is augmented as a probabilistic interval based on the convolution of the noise and delay. Third, we propose a fusion algorithm that admits the fused value with highest joint probability distribution of the intervals to tolerate corrupted measurements. Finally, we demonstrate the effectiveness of our method in a vehicle-platoon case study using extensive simulations and testbed experiments.

Recent emerging technologies such as artificial intelligence and machine learning have been promising enormous economic and societal benefits. While it is desirable to deploy these technologies to Internet-of-Things (IoT) infrastructures in many applications such as medical, energy, transportation, and industrial automation systems, such deployments present daunting challenges in performance, efficiency, and dependability of scaling-up IoT infrastructure, due to the ever-increasing number of edge devices, ever-increasing levels of device and system heterogeneity, and more stringent requirements of reliability, robustness, and security in mission-critical settings. This position paper elaborates the needs for a cross-layer and full hardware/software stack solution for the design and deployment of scalable, secure, and smart mission-critical IoT systems from four different perspectives and research fields. We present a review of recent studies on such issues and identify the potential challenges and gaps, based on which we highlight some important research directions and future works that can be conducted to tackle such challenges.

Cyber-Physical Systems (CPS) tightly couple information technology with physical processes, which rises new vulnerabilities such as physical attacks that are beyond conventional cyber attacks. Attackers may non-invasively compromise sensors and spoof the controller to perform unsafe actions. This issue is even emphasized with the increasing autonomy in CPS. While this fact has motivated many defense mechanisms against sensor attacks, a clear vision on the timing and usability (or the false alarm rate) of attack detection still remains elusive. Existing works tend to pursue an unachievable goal of minimizing the detection delay and false alarm rate at the same time, while there is a clear trade-off between the two metrics. Instead, we argue that attack detection should bias different metrics when a system sits in different states. For example, if the system is close to unsafe states, reducing the detection delay is preferable to lowering the false alarm rate, and vice versa. To achieve this, we make the following contributions. In this paper, we propose a real-time adaptive sensor attack detection framework. The framework can dynamically adapt the detection delay and false alarm rate so as to meet a detection deadline and improve the usability according to different system status. The core component of this framework is an attack detector that identifies anomalies based on a CUSUM algorithm through monitoring the cumulative sum of difference (or residuals) between the nominal (predicted) and observed sensor values. We augment this algorithm with a drift parameter that can govern the detection delay and false alarm. The second component is a behavior predictor that estimates nominal sensor values fed to the core component for calculating the residuals. The predictor uses a deep learning model that is offline extracted from sensor data through leveraging convolutional neural network (CNN) and recurrent neural network (RNN). The model relies on little knowledge of the system (e.g., dynamics), but uncovers and exploits both the local and complex long-term dependencies in multivariate sequential sensor measurements. The third component is a drift adaptor that estimates a detection deadline and then determines the drift parameter fed to the detector component for adjusting the detection delay and false alarms. Finally, we implement the proposed framework and validate it using realistic sensor data of automotive CPS to demonstrate its efficiency and efficacy.

As the connectivity of autonomous vehicles keeps growing, it is an accepted fact that they are even more vulnerable to malicious cyber-attacks. Recently, sensor spoofing has become an emerging attack that can compromise vehicle safety as vehicles are equipped with more sensors. Thus, it is critical to validate the sensor readings before utilizing them for future actions. In this paper, we develop HDAD, a hyperdimensional computing-based anomaly detection method. Hyperdimensional computing (HDC) is an emerging brain-inspired computing paradigm that mimics the brain cognition and leverages hyperdimensional vectors with fully distributed holographic representation and (pseudo)randomness. The key idea of HDAD is to use HDC to build encoder and decoder to reconstruct the sensor readings. The anomalous data typically have comparatively higher reconstruction errors than normal sensor readings. We explore three different metrics to measure the reconstruction error including mean squared error, mean absolute error, and cosine similarity. Using a real-world vehicle sensor reading dataset, we demonstrate the feasibility and efficacy of HDAD, opening the door for a new set of anomaly detection algorithm design.

Autonomous cyber-physical systems (CPS) are susceptible to non-invasive physical attacks such as sensor spoofing attacks that are beyond the classical cybersecurity domain. These attacks have motivated numerous research efforts on attack detection, but little attention on what to do after detecting an attack. The importance of attack recovery is emphasized by the need to mitigate the attack’s impact on a system and restore it to continue functioning. There are only a few works addressing attack recovery, but they all rely on prior knowledge of system dynamics. To overcome this limitation, we propose Recovery-by-Learning, a data-driven attack recovery framework that restores CPS from sensor attacks. The framework leverages natural redundancy among heterogeneous sensors and historical data for attack recovery. Specially, the framework consists of two major components: state predictor and data checkpointer. First, the predictor is triggered to estimate systems states after the detection of an attack. We propose a deep learning-based prediction model that exploits the temporal correlation among heterogeneous sensors. Second, the checkpointer executes when no attack is detected. We propose a double sliding window based checkpointing protocol to remove compromised data and keep trustful data as input to the state predictor. Third, we implement and evaluate the effectiveness of our framework using a realistic data set and a ground vehicle simulator. The results show that our method restores a system to continue functioning in presence of sensor attacks.

The increasing autonomy and connectivity in cyber-physical systems (CPS) come with new security vulnerabilities that are easily exploitable by malicious attackers to spoof a system to perform dangerous actions. While the vast majority of existing works focus on attack prevention and detection, the key question is “what to do after detecting an attack?”. This problem attracts fairly rare attention though its significance is emphasized by the need to mitigate or even eliminate attack impacts on a system. In this article, we study this attack response problem and propose novel real-time recovery for securing CPS. First, this work’s core component is a recovery control calculator using a Linear-Quadratic Regulator (LQR) with timing and safety constraints. This component can smoothly steer back a physical system under control to a target state set before a safe deadline and maintain the system state in the set once it is driven to it. We further propose an Alternating Direction Method of Multipliers (ADMM) based algorithm that can fast solve the LQR-based recovery problem. Second, supporting components for the attack recovery computation include a checkpointer, a state reconstructor, and a deadline estimator. To realize these components respectively, we propose (i) a sliding-window-based checkpointing protocol that governs sufficient trustworthy data, (ii) a state reconstruction approach that uses the checkpointed data to estimate the current system state, and (iii) a reachability-based approach to conservatively estimate a safe deadline. Finally, we implement our approach and demonstrate its effectiveness in dealing with totally 15 experimental scenarios which are designed based on 5 CPS simulators and 3 types of sensor attacks.

The increasing autonomy and connectivity have been transitioning automobiles to complex and open architectures that are vulnerable to malicious attacks beyond conventional cyber attacks. Attackers may non-invasively compromise sensors and spoof the controller to perform unsafe actions. This concern emphasizes the need to validate sensor data before acting on them. Unlike existing works, this paper exploits inherent redundancy among heterogeneous sensors for detecting anomalous sensor measurements. The redundancy is that multiple sensors simultaneously respond to the same physical phenomenon in a related fashion. Embedding the redundancy into a deep autoencoder, we propose an anomaly detector that learns a consistent pattern from vehicle sensor data in normal states and utilizes it as the nominal behavior for the detection. The proposed method is independent of the scarcity of anomalous data for training and the intensive calculation of pairwise correlation among senors as in existing works. Using a real-world data set collected from tens of vehicle sensors, we demonstrate the feasibility and efficacy of the proposed method.

Attack detection and recovery are fundamental elements for the operation of safe and resilient cyber-physical systems. Most of the literature focuses on attack-detection, while leaving attack-recovery as an open problem. In this paper, we propose novel attack-recovery control for securing cyber-physical systems. Our recovery control consists of new concepts required for a safe response to attacks, which includes the removal of poisoned data, the estimation of the current state, a prediction of the reachable states, and the online design of a new controller to recover the system. The synthesis of such recovery controllers for cyber-physical systems has barely investigated so far. To fill this void, we present a formal method-based approach to online compute a recovery control sequence that steers a system under an ongoing sensor attack from the current state to a target state such that no unsafe state is reachable on the way. The method solves a reach-avoid problem on a Linear Time-Invariant (LTI) model with the consideration of an error bound ε ≥ 0. The obtained recovery control is guaranteed to work on the original system if the behavioral difference between the LTI model and the system's plant dynamics is not larger than ε. Since a recovery control should be obtained and applied at the runtime of the system, in order to keep its computational time cost as low as possible, our approach firstly builds a linear programming restriction with the accordingly constrained safety and target specifications for the given reach-avoid problem, and then uses a linear programming solver to find a solution. To demonstrate the effectiveness of our method, we provide (a) the comparison to the previous work over 5 system models under 3 sensor attack scenarios: modification, delay, and reply; (b) a scalability analysis based on a scalable model to evaluate the performance of our method on large-scale systems.

With the fast proliferation of cloud computing, major cloud service providers, e.g., Amazon, Google, Facebook, etc., have been deploying more and more geographically distributed data centers to provide customers with better reliability and quality of services. A basic demand in such a geo-distributed data center system is to transfer bulk volumes of data from one data center to another. Geographic distribution and large delay-tolerance of such inter-data-center bulk data transfers provide cloud service providers opportunities to optimize the operating cost. Most existing studies on inter-data-center bulk data transfers focus on minimizing the network bandwidth cost. However, the energy-cost of the bulk data transfers, which also accounts for a large proportion of operating cost in the data centers, still remains unexplored. This is an important problem, especially in the multi-electricity-market environment, where the electricity price exhibits both spatial and temporal diversities. In this paper, we systematically study the problem of how to route and schedule inter-data-center bulk data transfers to minimize the energy-cost for geo-distributed data centers. We model this problem as a min-cost multi-commodity flow problem and develop an efficient two-stage optimization method to solve it. Extensive evaluations with real-life inter-data-center network and electricity prices show that our method brings significant energy-cost savings over existing bulk data transfer methods.

Demand response is considered as a valuable functionality of the power grid and its potential impacts continue expanding with grid modernization. Colocation data centers (simply called colocation) are recognized as a notably promising resource for demand response due to their high power demand and remarkable potential in demand management. A major challenge of colocation demand response is the split incentive, that is, colocation operators desire demand response for financial compensation while tenants may not embrace demand response due to lack of incentives. Another key challenge is caused by renewable energy co-located with data centers. Demand response mechanisms overlooking the uncertainty of renewable would cause much inefficiency in terms of energy saving and economic aspects. Existing work considers the two challenges separately in the context of data centers. By contrast, this work jointly addresses them and specially studies mechanism design for colocation data centers in presence of co-located renewable. We propose a hierarchical demand response scheme, which is based on a new two-level market mechanism that results in a win-win situation for both parties, i.e., tenants who choose to reduce power demand obtain financial rewards from the operator, while the operator receives financial compensation from the electric power company due to its tenants' demand reduction. At each demand response period, the colocation operator solicits bids (amount of energy reduction) from tenants and tenants who choose to participate responds to the operator with their bids. The proposed mechanism provably converges to a unique equilibrium solution, and at the equilibrium, neither the operator or tenants can improve their individual economic performance by changing their own strategies. Further, we present a stochastic optimization based algorithm, which uses predictions of the co-located renewable to determine the colocation operator's best strategy. At the equilibrium, the algorithm has a provable economic performance guarantee in terms of the prediction error. We finally evaluate the designed mechanism via detailed simulations and the results show the efficacy and validate the theoretical analysis for the mechanism.

With the large-scale deployment of Electric Vehicles (EVs), the unbalanced distribution of charging needs and random charging behaviors cause charging stations (CSs) congestion. This degrades EV drivers' quality of experience by extending charging waiting time and increasing charging fee. Thus, EV owners are facing a critical issue on how to decrease the cost of charging, which consists of two parts: charging duration and charging fee. A great deal of existing work is confined to finding CSs to optimize the two parts individually. However, it still remains unexplored how to jointly minimize charging duration and charging fee under an overall time limit (i.e., deadline) of a scheduled trip. The problem is the focus of this paper. First, we formulate this problem as a 0-1 Integer Linear Programming problem and show its NP-Hardness. Then, we propose an efficient distributed algorithm based on the Alternating Direction Method of Multipliers (ADMM). The algorithm decomposes the original problem into sub-problems that can be solved locally and in parallel between charging stations and the global coordinator. Finally, we carry out extensive simulations based on real-life transport network data, and the results show that the proposed approach brings significant cost savings over existing ones.

Cloud-based video streaming systems such as YouTube and Netflix are usually supported by the content delivery networks and data centers that can consume many megawatts of power. Most existing work independently studies the issues of improving quality of experience (QoE) for viewers and reducing the cost and emissions associated with the enormous energy usage of data centers. By contrast, this paper addresses them both, and jointly optimizes the QoE, the energy cost and emissions by intelligently allocating data center bandwidth among different client groups. Specially, we propose a distributed algorithm to achieve the optimal bandwidth allocation, given the prediction of future workload. The algorithm novelly decomposes the optimization process into separate ones, which are solved iteratively across data centers and clients. Further, the algorithm has robust performance guarantee in terms of the variance of the prediction error. We demonstrate its convergence and robustness by both proofs using theoretical analysis and validation based on trace-driven simulations. The results further show that the proposed algorithm converges very fast and achieves much better QoE-cost balance than existing approaches.

A smart factory generates vast amounts of data that require transmission via large-scale wireless networks. Thus, the reliability and real-time performance of large-scale wireless networks are essential for industrial production. A distributed data transmission scheme is suitable for large-scale networks, but is incapable of optimizing performance. By contrast, a centralized scheme relies on knowledge of global information and is hindered by scalability issues. To overcome these limitations, a hybrid scheme is needed. We propose a hierarchical data transmission framework that integrates the advantages of these schemes and makes a tradeoff among real-time performance, reliability, and scalability. The top level performs coarse-grained management to improve scalability and reliability by coordinating communication resources among subnetworks. The bottom level performs fine-grained management in each subnetwork, for which we propose an intrasubnetwork centralized scheduling algorithm to schedule periodic and aperiodic flows. We conduct both extensive simulations and realistic testbed experiments. The results indicate that our method has better schedulability and reduces packet loss by up to $22\%$ relative to existing methods.

Large-scale charging stations become indispensable infrastructure to support the rapid proliferation of electric vehicles. Their operation modes have drawn great attention from both academia and industry. One promising mode called park-and-charge has been recently introduced. This new mode allows customers to park their electric vehicles at a parking lot, where the vehicles are charged during the parking time. Several small-scale experiments, such as the V-Charge project and General Motors' E-Motor plant, have demonstrated its potential. A key enabler for deploying this mode to large-scale stations is effective and efficient charging load scheduling methods. Most existing works confine to the time-driven scheduling policy due to their sole focus on the charging service. Applying their solutions to the park-and-charge mode would jeopardize the unitization of charging resource or cause frequent charging mode switching. This inapplicability motivates us to explore the feasibility and benefits of exploiting the event-driven scheduling policy in park-and-charge systems. Further, to better characterize charging load in this mode, we propose to adopt a metered model, by which a system gains value in proportion to the served charging demand. To be specific, the objective of this paper is to carry out both theoretical and experimental analysis for event-driven algorithms adapted to this metered model. We leverage both the competitive analysis and resource augmentation to demonstrate the non-constant and constant performance bounds for the earliest-deadline-first and highest-value-first algorithms respectively. Moreover, we provide a stronger theoretical result, i.e., the performance bound for the whole class of work-conserving scheduling algorithms. Through extensive simulations, we validate the proposed theoretical results and further provide interesting findings from the in-depth analysis of the simulation results.

The fast proliferation of cloud computing promotes the rapid development of large-scale commercial data centers. Tens or even hundreds of geographically distributed data centers have been deployed for better reliability and quality of services. This brings huge energy consumption for data centers. Previous research has proved that the geographical load balancing technique can achieve significant energy cost savings for geographically distributed data centers. However, existing methods for geographical load balancing often assume data centers with homogeneous servers, and workloads with single-dimension or uniform resource demands. This is an over-simplification in reality, especially when modern data centers are typically constructed from a variety of server classes. In this paper, we systematically study the problem of job scheduling for geographically distributed data centers to embrace the heterogeneity of underlying platforms and workloads. We develop a novel distributed algorithm to solve the problem efficiently based on the alternating direction method of multipliers. Extensive evaluations based on real-life data center topology, traffic traces, and electricity price data show high efficiency and efficacy of our method.

Video streaming systems such as YouTube and Netflix are usually supported by the content delivery networks and datacenters that can consume many megawatts of power. Most existing works independently study the issues of improving quality of experience (QoE) for viewers and reducing the cost and emissions associated with the enormous energy usage of datacenters. By contrast, this paper addresses them both, and jointly optimizes the QoE, the energy cost and emissions by intelligently allocating datacenter bandwidth among different client groups. Specially, we propose a distributed algorithm for achieving the optimal bandwidth allocation. The algorithm novelly decomposes the optimization process into separate ones, which are solved iteratively across datacenters and clients. We demonstrate its convergence by both theoretical proof and experimental validation. The experimental results show that the proposed algorithm converges very fast and achieves much better QoE-cost balance than existing approaches.

Component-based software systems with real-time requirements are often scheduled using processor reservation techniques. Such techniques have mainly evolved around hard real-time systems in which worst-case resource demands are considered for the reservations. In soft real-time systems, reserv- ing the processors based on the worst-case demands results in unnecessary over-allocations. In this paper, targeting soft real-time systems running on multiprocessor platforms, we focus on components for which processor demand varies during run-time. We propose a feedback scheduling framework where processor reservations are used for scheduling components. The reservation bandwidths as well as the reservation periods are adapted using MIMO LQR controllers. We provide an allocation mechanism for distributing components over processors. The proposed framework is implemented in the TrueTime simulation tool for system identification. We use a case study to investigate the performance of our framework in the simulation tool. Finally, the framework is implemented in the Linux kernel for practical evaluations. The evaluation results suggest that the framework can efficiently adapt the reservation parameters during run-time by imposing negligible overhead.

Demand response is an important feature and functionality of the future smart grid. Electric vehicles are recognized as a particularly promising resource for demand response given their high charging demand and flexibility in demand management. Recently, researchers begun to apply market-based solutions to electric vehicle demand response. A clear vision, however, remains elusive because existing works overlook three key issues. (i) The hierarchy among electric vehicles (EVs), charging stations, and electric power companies (EPCs). Previous works assume direct interaction between EVs and EPCs and thus confine to single-level market designs. The designed mechanisms are inapplicable here due to ignoring the role of charging stations in the hierarchy. (ii) Temporal aspects of charging loads. Solely focusing on economic aspects makes significant demand reduction, but electric vehicles would end up with little allocated power due to overlooking their temporal constraints. (iii) Renewable generation co-located with charging stations. Market mechanisms that overlook the uncertainty of renewable would cause much inefficiency in terms of both the economic and temporal aspects. To address these issues, we study a new demand response scheme, i.e, hierarchical demand response for electric vehicles via charging stations. We propose that two-level marketing is suitable to this hierarchical scheme, and design a distributed market mechanism that is compatible with both the economic and temporal aspects of electric vehicle demand response. The market mechanism has a hierarchical decision-making structure by which the charging station leads the market and electric vehicles follow and respond to its actions. An appealing feature of the mechanism is the provable convergence to a unique equilibrium solution. At the equilibrium, neither the charging station or electric vehicles can improve their individual economic and/or temporal performance by changing their own strategies. Furthermore, we present a stochastic optimization based algorithm to optimize economic performance for the charging station at the equilibrium, given the predictions of the co-located renewable generation. The algorithm has provable robust performance guarantee in terms of the variance of the prediction errors. We finally evaluate the designed mechanism via detailed simulations. The results show the efficacy and validate the theoretical analysis for the mechanism.

The growing awareness about global climate change has boosted the need to mitigate greenhouse gas emissions from existing power systems and spurred efforts to accelerate the integration of renewable energy sources (e.g. wind and solar power) into the electrical grid. A fundamental difficulty here is that renewable energy sources are usually of high variability. The electrical grid must absorb this variability through employing many additional operations (e.g., operating reserves, energy storage), which will largely raise the cost of electricity from renewable energy sources. To make it affordable, numerous advancements in technologies and methods for the smart grid are required. In this paper, we will confine ourselves to one of them: how to plan the construction of wind farms with high capacity and low variability locally and distributedly. We first study the characteristics of both wind resources and wind turbines and present a more accurate wind power evaluation method based on Gaussian Regression. Then, we analyze a trade-off between wind power's quantity and quality and propose an approach to optimally combine different types of wind turbines to balance the trade-off for a specific site. Finally, we explore geographical diversity among different sites and develop an extended approach that jointly optimizes the combination of sites and turbine types. Extensive experiments using the realistic historical wind resource data are conducted for either of the local and distributed case. Encouraging results are shown for the proposed approaches and some interesting insights are also provided.

To reduce cost and emission, modern datacenter operators are beginning to incorporate green energy sources into datacenters' power supply. To improve service availability, they also back up datacenters using traditional (usually brown) energy sources. However, challenge arises due to distinct characteristics of energy sources used for different goals. How to select optimal energy sources and plan their capacity for datacenters to meet cost, emission and service availability requirement remains an open research problem. In this extended abstract, we briefly describe recent work in [4], which provides a holistic solution to address this problem. In [4], we present GreenPlanning, a framework to strike a judicious balance among multiple energy sources, the electrical grid and energy storage devices for a datacenter in terms of cost, emission, and service availability. GreenPlanning explores different features and operations of both green and traditional energy sources available to datacenters. The framework minimizes the lifetime total cost including both capital and operational cost for a datacenter. We conduct extensive experiments to evaluate GreenPlanning with real-life computational workload and meteorological data traces. Results demonstrate that GreenPlanning can reduce the lifetime total cost and emission by more than 50% compared to traditional configurations without integration of green energy, while still meeting service availability requirement.

The need to reduce greenhouse gases from our current power systems accelerates the integration of renewable energy sources (for example, wind and solar power). A fundamental difficulty is that renewable energy is usually of high variability. Numerous advancements in technologies and methods for the smart grid are required to mitigate and absorb this variability. In this paper, we focus on one of them: how to plan wind farms with high capacity and low variability locally and distributedly. First, we study the characteristics of both wind resource and wind turbines and propose a novel wind power estimation method based on Gaussian regression. The experimental result shows that our method achieves a more accurate estimation compared to other ones and has a nearly zero error for most of the turbine types. Then, we analyze a tradeoff between wind power's quantity and quality for large-scale wind farms, and find that there is an optimal turbine type for each location as to either the quantity or the quality. We propose an approach to optimally combine different types of wind turbines to balance the tradeoff. Finally, we explore geographical diversity among different locations and develop an extended approach that jointly optimizes the combination of locations and turbine types. Besides applying to plan new wind farms, we also discuss how to adapt the two approaches to decide an upgrade plan for a wind farm and a network of wind farms, respectively. We conduct extensive experiments using two different wind resource data traces for both local and distributed cases. The result shows that the proposed approaches significantly outperform those approaches using a single turbine type and those separately optimizing locations and turbine types. We also provide interesting insights about the quantity-quality balancing.

While much work has addressed energy minimizing problem of real-time sequential tasks, little has been done for the parallel real-time task case. In this paper, based on level-packing, we study energy minimization problem for parallel task systems with discrete operation modes and under timing constraints. For tasks with fixed (variable) parallel degrees, we first formulate the problem as a 0-1 Integer Linear Program (0-1 ILP), and then propose a polynomial-time complexity two-step (three-step) heuristic to determine task schedule and frequency assignment (and the task parallel degree). Our simulation result shows that the heuristics consume nearly the same energy as do 0-1 ILPs.

While much work has addressed the energy-efficient scheduling problem for uniprocessor or multiprocessor systems, little has been done for multicore systems. We study the multicore architecture with a fixed number of cores partitioned into clusters (or islands), on each of which all cores operate at a common frequency. We develop algorithms to determine a schedule for real-time tasks to minimize the energy consumption under the timing and operating frequency constraints. As technical contributions, we first show that the optimal frequencies resulting in the minimum energy consumption for each island is not dependent on the workload mapped but the number of cores and leakage power on the island, when not considering the timing constraint. Then for systems with timing constraints, we present a polynomial algorithm which derives the minimum energy consumption for a given task partition. Finally, we develop an efficient algorithm to determine the number of active islands, task partition and frequency assignment. Our simulation result shows that our approach significantly outperforms the related approaches in terms of energy saving.

Energy conservation is an important issue in the design of embedded systems. Dynamic Voltage Scaling (DVS) and Dynamic Power Management (DPM) are two widely used techniques for saving energy in such systems. In this paper, we address the problem of minimizing multi-resource energy consumption concerning both CPU and devices. A system is assumed to contain a fixed number of real-time tasks scheduled to run on a DVS-enabled processor, and a fixed number of off-chip devices used by the tasks during their executions. We will study the non-trivial time and energy overhead of device state transitions between active and sleep states. Our goal is to find optimal schedules providing not only the execution order and CPU frequencies of tasks, but also the time points for device state transitions. We adopt the frame-based real-time task model, and develop optimization algorithms based on 0-1 Integer Non-Linear Programming (0-1 INLP) for different system configurations. Simulation results indicate that our approach can significantly outperform existing techniques in terms of energy savings.