Biblio

Cyber-physical systems are subject to natural uncertainties and sensor noise that can be amplified/attenuated due to feedback. In this work, we want to leverage these properties in order to define the inherent differential privacy of feedback-control systems without the addition of an external differential privacy noise. If larger levels of privacy are required, we introduce a methodology to add an external differential privacy mechanism that injects the minimum amount of noise that is needed. On the other hand, we show how the combination of inherent and external noise affects system security in terms of the impact that integrity attacks can impose over the system while remaining undetected. We formulate a bilevel optimization problem to redesign the control parameters in order to minimize the attack impact for a desired level of inherent privacy.

Statistical structure learning (SSL)-based approaches have been employed in the recent years to detect different types of anomalies in a variety of cyber-physical systems (CPS). Although these approaches outperform conventional methods in the literature, their computational complexity, need for large number of measurements and centralized computations have limited their applicability to large-scale networks. In this work, we propose a distributed, multi-agent maximum likelihood (ML) approach to detect anomalies in smart grid applications aiming at reducing computational complexity, as well as preserving data privacy among different players in the network. The proposed multi-agent detector breaks the original ML problem into several local (smaller) ML optimization problems coupled by the alternating direction method of multipliers (ADMM). Then, these local ML problems are solved by their corresponding agents, eventually resulting in the construction of the global solution (network's information matrix). The numerical results obtained from two IEEE test (power transmission) systems confirm the accuracy and efficiency of the proposed approach for anomaly detection.

Cyber-Physical Systems (CPSs) play an increasingly significant role in many critical applications. These valuable applications attract various sophisticated attacks. This paper considers a stealthy estimation attack, which aims to modify the state estimation of the CPSs. The intelligent attackers can learn defense strategies and use clandestine attack strategies to avoid detection. To address the issue, we design a Chi-square detector in a Digital Twin (DT), which is an online digital model of the physical system. We use a Signaling Game with Evidence (SGE) to find the optimal attack and defense strategies. Our analytical results show that the proposed defense strategies can mitigate the impact of the attack on the physical estimation and guarantee the stability of the CPSs. Finally, we use an illustrative application to evaluate the performance of the proposed framework.

Adversarial examples are known as carefully perturbed images fooling image classifiers. We propose a geometric framework to generate adversarial examples in one of the most challenging black-box settings where the adversary can only generate a small number of queries, each of them returning the top-1 label of the classifier. Our framework is based on the observation that the decision boundary of deep networks usually has a small mean curvature in the vicinity of data samples. We propose an effective iterative algorithm to generate query-efficient black-box perturbations with small p norms which is confirmed via experimental evaluations on state-of-the-art natural image classifiers. Moreover, for p=2, we theoretically show that our algorithm actually converges to the minimal perturbation when the curvature of the decision boundary is bounded. We also obtain the optimal distribution of the queries over the iterations of the algorithm. Finally, experimental results confirm that our principled black-box attack algorithm performs better than state-of-the-art algorithms as it generates smaller perturbations with a reduced number of queries.

Style transfer produces a transferred image which is a rendering of a content image in the manner of a style image. We seek to understand how to improve style transfer.To do so requires quantitative evaluation procedures, but current evaluation is qualitative, mostly involving user studies. We describe a novel quantitative evaluation procedure. Our procedure relies on two statistics: the Effectiveness (E) statistic measures the extent that a given style has been transferred to the target, and the Coherence (C) statistic measures the extent to which the original image's content is preserved. Our statistics are calibrated to human preference: targets with larger values of E and C will reliably be preferred by human subjects in comparisons of style and content, respectively.We use these statistics to investigate relative performance of a number of Neural Style Transfer (NST) methods, revealing a number of intriguing properties. Admissible methods lie on a Pareto frontier (i.e. improving E reduces C, or vice versa). Three methods are admissible: Universal style transfer produces very good C but weak E; modifying the optimization used for Gatys' loss produces a method with strong E and strong C; and a modified cross-layer method has slightly better E at strong cost in C. While the histogram loss improves the E statistics of Gatys' method, it does not make the method admissible. Surprisingly, style weights have relatively little effect in improving EC scores, and most variability in transfer is explained by the style itself (meaning experimenters can be misguided by selecting styles). Our GitHub Link is available1.

This paper studies the data-driven stealthy actuator attack against cyber-physical systems. The objective of the attacker is to add a certain bias to the output while keeping the detection rate of the χ2 detector less than a certain value. With the historical input and output data, the parameters of the system are estimated and the attack signal is the solution of a convex optimization problem constructed with the estimated parameters. The extension to the case of arbitrary detectors is also discussed. A numerical example is given to verify the effectiveness of the attack.

With rapid development of deep integration of computation, control, and communication, Cyber-Physical Systems (CPSs) play an important role in industrial processes. Combined with the technology of fog computing, CPSs can outsource their complicated computation to the fog layer, which in turn, may bring security threats with regard to data privacy. To protect data privacy in a control framework, this paper investigate observer-based secure control problem towards fog computing aided CPSs (FCA-CPSs) by utilizing data perturbation method. Firstly, security inputs are designed to encrypt the transmitted states to realize specific confidentiality level. Then, sufficient conditions are established to ensure the stability of considered FCA-CPSs. Finally, a numerical example is provided to illustrate the effectiveness of the secure estimation scheme.

This paper considers consensus-based distributed Kalman filtering subject to data-falsification attack, where Byzantine agents share manipulated data with their neighboring agents. The attack is assumed to be coordinated among the Byzantine agents and follows a linear model. The goal of the Byzantine agents is to maximize the network-wide estimation error while evading false-data detectors at honest agents. To that end, we propose a joint selection of Byzantine agents and covariance matrices of attack sequences to maximize the network-wide estimation error subject to constraints on stealthiness and the number of Byzantine agents. The attack strategy is then obtained by employing block-coordinate descent method via Boolean relaxation and backward stepwise based subset selection method. Numerical results show the efficiency of the proposed attack strategy in comparison with other naive and uncoordinated attacks.

We consider distributed Kalman filter for dynamic state estimation over wireless sensor networks. It is promising but challenging when network is under cyber attacks. Since the information exchange between nodes, the malicious attacks quickly spread across the entire network, which causing large measurement errors and even to the collapse of sensor networks. Aiming at the malicious network attack, a trust-based distributed processing frame is proposed. Which allows neighbor nodes to exchange information, and a series of trusted nodes are found using truth discovery. As a demonstration, distributed Cooperative Localization is considered, and numerical results are provided to evaluate the performance of the proposed approach by considering random, false data injection and replay attacks.

SPARQL is a standard query language for knowledge graphs (KGs). However, it is hard to find correct answer if KGs are incomplete or incorrect. Knowledge graph embedding (KGE) enables answering queries on such KGs by inferring unknown knowledge and removing incorrect knowledge. Hence, our long-term goal in this line of research is to propose a new framework that integrates KGE and SPARQL, which opens various research problems to be addressed. In this paper, we solve one of the most critical problems, that is, optimizing the performance of nearest neighbor (NN) search. In our evaluations, we demonstrate that the search time of state-of-the-art NN search algorithms is improved by 40% without sacrificing answer accuracy.

In fault diagnosis of industrial process, there are usually more than one variable that are faulty. When multiple faults occur, the generalized reconstruction-based contribution can be helpful while traditional RBC may make mistakes. Due to the correlation between the variables, these faults usually propagate to other normal variables, which is called smearing effect. Thus, it is helpful to consider the pervious fault diagnosis results. In this paper, a data-driven fault diagnosis method which is based on generalized RBC and bayesian decision is presented. This method combines multi-dimensional RBC and bayesian decision. The proposed method improves the diagnosis capability of multiple and minor faults with greater noise. A numerical simulation example is given to show the effectiveness and superiority of the proposed method.

Principal Component Analysis (PCA) is a basic method of fault diagnosis based on multivariate statistical analysis. It utilizes the linear correlation between multiple process variables to implement process fault diagnosis and has been widely used. Traditional PCA fault diagnosis ignores the impact of faults with different magnitudes on detection accuracy. Based on a variety of data processing methods, this paper proposes a multi-level and multi-granularity principal component analysis method to make the detection results more accurate.

Gaussian random attacks that jointly minimize the amount of information obtained by the operator from the grid and the probability of attack detection are presented. The construction of the attack is posed as an optimization problem with a utility function that captures two effects: firstly, minimizing the mutual information between the measurements and the state variables; secondly, minimizing the probability of attack detection via the Kullback-Leibler (KL) divergence between the distribution of the measurements with an attack and the distribution of the measurements without an attack. Additionally, a lower bound on the utility function achieved by the attacks constructed with imperfect knowledge of the second order statistics of the state variables is obtained. The performance of the attack construction using the sample covariance matrix of the state variables is numerically evaluated. The above results are tested in the IEEE 30-Bus test system.

We consider the problem of enabling robust range estimation of eigenvalue decomposition (EVD) algorithm for a reliable fixed-point design. The simplicity of fixed-point circuitry has always been so tempting to implement EVD algorithms in fixed-point arithmetic. Working towards an effective fixed-point design, integer bit-width allocation is a significant step which has a crucial impact on accuracy and hardware efficiency. This paper investigates the shortcomings of the existing range estimation methods while deriving bounds for the variables of the EVD algorithm. In light of the circumstances, we introduce a range estimation approach based on vector and matrix norm properties together with a scaling procedure that maintains all the assets of an analytical method. The method could derive robust and tight bounds for the variables of EVD algorithm. The bounds derived using the proposed approach remain same for any input matrix and are also independent of the number of iterations or size of the problem. Some benchmark hyperspectral data sets have been used to evaluate the efficiency of the proposed technique. It was found that by the proposed range estimation approach, all the variables generated during the computation of Jacobi EVD is bounded within ±1.

Lithium Ion batteries usually degrade to an unacceptable capacity level after hundreds or even thousands of cycles. The continuously observed capacity fade data over time and their internal structure can be informative for constructing capacity fade models. This paper applies a mean-covariance decomposition modeling method to analyze the capacity fade data. The proposed approach directly examines the variances and correlations in data of interest and express the correlation matrix in hyper-spherical coordinates using angles and trigonometric functions. The proposed method is applied to model and predict key batteries performance metrics using testing data under various testing conditions.

Standard classification procedures of both data mining and multivariate statistics are sensitive to the presence of outlying values. In this paper, we propose new algorithms for computing regularized versions of linear discriminant analysis for data with small sample sizes in each group. Further, we propose a highly robust version of a regularized linear discriminant analysis. The new method denoted as MWCD-L2-LDA is based on the idea of implicit weights assigned to individual observations, inspired by the minimum weighted covariance determinant estimator. Classification performance of the new method is illustrated on a detailed analysis of our pilot study of authentication methods on computers, using individual typing characteristics by means of keystroke dynamics.

In adaptive processing applications, the design of the adaptive filter requires estimation of the unknown interference-plus-noise covariance matrix from secondary training data. The presence of outliers in the training data can severely degrade the performance of adaptive processing. By exploiting the sparse prior of the outliers, a Bayesian framework to develop a computationally efficient outlier-resistant adaptive filter based on sparse Bayesian learning (SBL) is proposed. The expectation-maximisation (EM) algorithm is used therein to obtain a maximum a posteriori (MAP) estimate of the interference-plus-noise covariance matrix. Numerical simulations demonstrate the superiority of the proposed method over existing methods.

Wireless channel reciprocity can be successfully exploited as a common source of randomness for the generation of a secret key by two legitimate users willing to achieve confidential communications over a public channel. This paper presents an analytical framework to investigate the theoretical limits of secret-key generation when wireless multi-dimensional Gaussian channels are used as source of randomness. The intrinsic secrecy content of wide-sense stationary wireless channels in frequency, time and spatial domains is derived through asymptotic analysis as the number of observations in a given domain tends to infinity. Some significant case studies are presented where single and multiple antenna eavesdroppers are considered. In the numerical results, the role of signal-to-noise ratio, spatial correlation, frequency and time selectivity is investigated.