Biblio

Neural networks with deep architectures have been used to construct state-of-the-art classifiers that can match human level accuracy in areas such as image classification. However, many of these classifiers can be fooled by examples slightly modified from their original forms. In this work, we propose a novel approach for generating adversarial examples that makes use of only attribution information of the features and perturbs only features that are highly influential to the output of the classifier. We call this approach Attribution Based Adversarial Generation (ABAG). To demonstrate the effectiveness of this approach, three somewhat arbitrary algorithms are proposed and examined. In the first algorithm all non-zero attributions are utilized and associated features perturbed; in the second algorithm only the top-n most positive and top-n most negative attributions are used and corresponding features perturbed; and in the third algorithm the level of perturbation is increased in an iterative manner until an adversarial example is discovered. All of the three algorithms are implemented and experiments are performed on the well-known MNIST dataset. Experiment results show that adversarial examples can be generated very efficiently, and thus prove the validity and efficacy of ABAG - utilizing attributions for the generation of adversarial examples. Furthermore, as shown by examples, ABAG can be adapted to provides a systematic searching approach to generate adversarial examples by perturbing a minimum amount of features.

The intelligent electronic devices widely deployed across the distribution network are inevitably making the feeder automation (FA) system more vulnerable to cyber-attacks, which would lead to disastrous socio-economic impacts. This paper proposes a three-stage game-theoretic framework that the defender allocates limited security resources to minimize the economic impacts on FA system while the attacker deploys limited attack resources to maximize the corresponding impacts. Meanwhile, the probability of successful attack is calculated based on the Bayesian attack graph, and a fault-tolerant location technique for centralized FA system is elaborately considered during analysis. The proposed game-theoretic framework is converted into a two-level zero-sum game model and solved by the particle swarm optimization (PSO) combined with a generalized reduced gradient algorithm. Finally, the proposed model is validated on distribution network for RBTS bus 2.

We leverage deep learning algorithms on various user behavioral information gathered from end-user devices to classify a subject of interest. In spite of the ability of these techniques to counter spoofing threats, they are vulnerable to adversarial learning attacks, where an attacker adds adversarial noise to the input samples to fool the classifier into false acceptance. Recently, a handful of mature techniques like Fast Gradient Sign Method (FGSM) have been proposed to aid white-box attacks, where an attacker has a complete knowledge of the machine learning model. On the contrary, we exploit a black-box attack to a behavioral biometric system based on gait patterns, by using FGSM and training a shadow model that mimics the target system. The attacker has limited knowledge on the target model and no knowledge of the real user being authenticated, but induces a false acceptance in authentication. Our goal is to understand the feasibility of a black-box attack and to what extent FGSM on shadow models would contribute to its success. Our results manifest that the performance of FGSM highly depends on the quality of the shadow model, which is in turn impacted by key factors including the number of queries allowed by the target system in order to train the shadow model. Our experimentation results have revealed strong relationships between the shadow model and FGSM performance, as well as the effect of the number of FGSM iterations used to create an attack instance. These insights also shed light on deep-learning algorithms' model shareability that can be exploited to launch a successful attack.

Increasing deep neural networks are applied in solving graph evolved tasks, such as node classification and link prediction. However, the vulnerability of deep models can be revealed using carefully crafted adversarial examples generated by various adversarial attack methods. To explore this security problem, we define the link prediction adversarial attack problem and put forward a novel iterative gradient attack (IGA) strategy using the gradient information in the trained graph autoencoder (GAE) model. Not surprisingly, GAE can be fooled by an adversarial graph with a few links perturbed on the clean one. The results on comprehensive experiments of different real-world graphs indicate that most deep models and even the state-of-the-art link prediction algorithms cannot escape the adversarial attack, such as GAE. We can benefit the attack as an efficient privacy protection tool from the link prediction of unknown violations. On the other hand, the adversarial attack is a robust evaluation metric for current link prediction algorithms of their defensibility.

The article deals with the mathematical model for information security controls optimization and evaluation of the information security systems effectiveness. Distinctive features of APT attacks are given. The generalized efficiency criterion in which both the requirements of the return of security investment maximization and the return on attack minimization are simultaneously met. The generalized reduced gradient method for solving the optimization of the objective function based on formulated efficiency criterion is proposed.

This paper investigates the impact of authentication on effective capacity (EC) of an underwater acoustic (UWA) channel. Specifically, the UWA channel is under impersonation attack by a malicious node (Eve) present in the close vicinity of the legitimate node pair (Alice and Bob); Eve tries to inject its malicious data into the system by making Bob believe that she is indeed Alice. To thwart the impersonation attack by Eve, Bob utilizes the distance of the transmit node as the feature/fingerprint to carry out feature-based authentication at the physical layer. Due to authentication at Bob, due to lack of channel knowledge at the transmit node (Alice or Eve), and due to the threshold-based decoding error model, the relevant dynamics of the considered system could be modelled by a Markov chain (MC). Thus, we compute the state-transition probabilities of the MC, and the moment generating function for the service process corresponding to each state. This enables us to derive a closed-form expression of the EC in terms of authentication parameters. Furthermore, we compute the optimal transmission rate (at Alice) through gradient-descent (GD) technique and artificial neural network (ANN) method. Simulation results show that the EC decreases under severe authentication constraints (i.e., more false alarms and more transmissions by Eve). Simulation results also reveal that the (optimal transmission rate) performance of the ANN technique is quite close to that of the GTJ method.

This work uses adversarial perturbations to enhance deepfake images and fool common deepfake detectors. We created adversarial perturbations using the Fast Gradient Sign Method and the Carlini and Wagner L2 norm attack in both blackbox and whitebox settings. Detectors achieved over 95% accuracy on unperturbed deepfakes, but less than 27% accuracy on perturbed deepfakes. We also explore two improvements to deep-fake detectors: (i) Lipschitz regularization, and (ii) Deep Image Prior (DIP). Lipschitz regularization constrains the gradient of the detector with respect to the input in order to increase robustness to input perturbations. The DIP defense removes perturbations using generative convolutional neural networks in an unsupervised manner. Regularization improved the detection of perturbed deepfakes on average, including a 10% accuracy boost in the blackbox case. The DIP defense achieved 95% accuracy on perturbed deepfakes that fooled the original detector while retaining 98% accuracy in other cases on a 100 image subsample.

Federated learning (FL), as a recent advance of distributed machine learning, is capable of learning a model over the network without directly accessing the client's raw data. Nevertheless, the clients' sensitive information can still be exposed to adversaries via differential attacks on messages exchanged between the parameter server and clients. In this paper, we consider the widely used federating averaging (FedAvg) algorithm and propose to enhance the data privacy by the differential privacy (DP) technique, which obfuscates the exchanged messages by properly adding Gaussian noise. We analytically show that the proposed secure FedAvg algorithm maintains an O(l/T) convergence rate, where T is the total number of stochastic gradient descent (SGD) updates for local model parameters. Moreover, we demonstrate how various algorithm parameters can impact on the algorithm communication efficiency. Experiment results are presented to justify the obtained analytical results on the performance of the proposed algorithm in terms of testing accuracy.

In this paper, we apply machine learning to distributed private data owned by multiple data owners, entities with access to non-overlapping training datasets. We use noisy, differentially-private gradients to minimize the fitness cost of the machine learning model using stochastic gradient descent. We quantify the quality of the trained model, using the fitness cost, as a function of privacy budget and size of the distributed datasets to capture the trade-off between privacy and utility in machine learning. This way, we can predict the outcome of collaboration among privacy-aware data owners prior to executing potentially computationally-expensive machine learning algorithms. Particularly, we show that the difference between the fitness of the trained machine learning model using differentially-private gradient queries and the fitness of the trained machine model in the absence of any privacy concerns is inversely proportional to the size of the training datasets squared and the privacy budget squared. We successfully validate the performance prediction with the actual performance of the proposed privacy-aware learning algorithms, applied to: financial datasets for determining interest rates of loans using regression; and detecting credit card frauds using support vector machines.

In this paper, differential privacy protection method is applied to matrix factorization method that used to solve the recommendation problem. For centralized recommendation scenarios, a collaborative filtering recommendation model based on matrix factorization is established, and a matrix factorization mechanism satisfying ε-differential privacy is proposed. Firstly, the potential characteristic matrix of users and projects is constructed. Secondly, noise is added to the matrix by the method of target disturbance, which satisfies the differential privacy constraint, then the noise matrix factorization model is obtained. The parameters of the model are obtained by the stochastic gradient descent algorithm. Finally, the differential privacy matrix factorization model is used for score prediction. The effectiveness of the algorithm is evaluated on the public datasets including Movielens and Netflix. The experimental results show that compared with the existing typical recommendation methods, the new matrix factorization method with privacy protection can recommend within a certain range of recommendation accuracy loss while protecting the users' privacy information.

This work presents the results of the Multi-Robot System designing that works on the basis of Swarm Intelligence models and is used to search for optimal solutions. The process of searching for optimal solutions is performed based on a field of gradient vectors that can be generated by ionizing radiation sources, radio-electro-magnetic devices, temperature generating sources, etc. The concept of the operation System is based on the distribution in the search space of a multitude of Mobile Robots that form a Mesh network between them. Each Mobile Robot has a set of ultrasonic sensors for excluding the collisions with obstacles, two sensors for identifying the gradient vector of the analyzed field, resources for wireless storage, processing and communication. The direction of the Mobile Robot movement is determined by the rotational speed of two DC motors which is calculated based on the models of Artificial Neural Networks. Gradient vectors generated by all Mobile Robots in the system structure are used to calculate the movement direction.

Transferring the style of an image is a fundamental problem in computer vision. Which extracts the features of a context image and a style image, then fixes them to produce a new image with features of the both two input images. In this paper, we introduce an artificial system to separate and recombine the content and style of arbitrary images, providing a neural algorithm for the creation of artistic images. We use a pre-trained deep convolutional neural network VGG19 to extract the feature map of the input style image and context image. Then we define a loss function that captures the difference between the output image and the two input images. We use the gradient descent algorithm to update the output image to minimize the loss function. Experiment results show the feasibility of the method.

Software Defined Networking (SDN) provides opportunities for flexible and dynamic traffic engineering. However, in current SDN systems, routing strategies are based on traditional mechanisms which lack in real-time modification and less efficient resource utilization. To overcome these limitations, deep learning is used in this paper to improve the routing computation in SDN. This paper proposes Convolutional Deep Reinforcement Learning (CoDRL) model which is based on deep reinforcement learning agent for routing optimization in SDN to minimize the mean network delay and packet loss rate. The CoDRL model consists of Deep Deterministic Policy Gradients (DDPG) deep agent coupled with Convolution layer. The proposed model tends to automatically adapts the dynamic packet routing using network data obtained through the SDN controller, and provides the routing configuration that attempts to reduce network congestion and minimize the mean network delay. Hence, the proposed deep agent exhibits good convergence towards providing routing configurations that improves the network performance.

In socially assistive robotics, an important research area is the development of adaptation techniques and their effect on human-robot interaction. We present a meta-learning based policy gradient method for addressing the problem of adaptation in human-robot interaction and also investigate its role as a mechanism for trust modelling. By building an escape room scenario in mixed reality with a robot, we test our hypothesis that bi-directional trust can be influenced by different adaptation algorithms. We found that our proposed model increased the perceived trustworthiness of the robot and influenced the dynamics of gaining human's trust. Additionally, participants evaluated that the robot perceived them as more trustworthy during the interactions with the meta-learning based adaptation compared to the previously studied statistical adaptation model.

Smart Grid cyber-security sounds to be a critical issue, because of widespread development of information technology. To achieve secure and reliable operation, the complexity of human automation interaction (HAI) necessitates more sophisticated and intelligent methodologies. In this paper, an adaptive autonomy fuzzy expert system is developed using gradient descent algorithm to determine the Level of Automation (LOA), based on the changing of Performance Shaping Factors (PSF). These PSFs indicate the effects of environmental conditions on the performance of HAI. The major advantage of this method is that the fuzzy rule or membership function can be learnt without changing the form of the fuzzy rule in conventional fuzzy control. Because of data shortage, Leave-One-Out Cross-Validation (LOOCV) technique is applied for assessing how the results of proposed system generalizes to the new contingency situations. The expert system database is extracted from superior experts' judgments. In order to regard the importance of each PSF, weighted rules are also considered. In addition, some new environmental conditions are introduced that has not been seen before. Nine scenarios are discussed to reveal the performance of the proposed system. Results confirm that the presented fuzzy expert system can effectively calculates the proper LOA even in the new contingency situations.

We propose a serverless computing mechanism for distributed computation based on polar codes. Serverless computing is an emerging cloud based computation model that lets users run their functions on the cloud without provisioning or managing servers. Our proposed approach is a hybrid computing framework that carries out computationally expensive tasks such as linear algebraic operations involving large-scale data using serverless computing and does the rest of the processing locally. We address the limitations and reliability issues of serverless platforms such as straggling workers using coding theory, drawing ideas from recent literature on coded computation. The proposed mechanism uses polar codes to ensure straggler-resilience in a computationally effective manner. We provide extensive evidence showing polar codes outperform other coding methods. We have designed a sequential decoder specifically for polar codes in erasure channels with full-precision input and outputs. In addition, we have extended the proposed method to the matrix multiplication case where both matrices being multiplied are coded. The proposed coded computation scheme is implemented for AWS Lambda. Experiment results are presented where the performance of the proposed coded computation technique is tested in optimization via gradient descent. Finally, we introduce the idea of partial polarization which reduces the computational burden of encoding and decoding at the expense of straggler-resilience.

Deep learning is a highly effective machine learning technique for large-scale problems. The optimization of nonconvex functions in deep learning literature is typically restricted to the class of first-order algorithms. These methods rely on gradient information because of the computational complexity associated with the second derivative Hessian matrix inversion and the memory storage required in large scale data problems. The reward for using second derivative information is that the methods can result in improved convergence properties for problems typically found in a non-convex setting such as saddle points and local minima. In this paper we introduce TRMinATR - an algorithm based on the limited memory BFGS quasi-Newton method using trust region - as an alternative to gradient descent methods. TRMinATR bridges the disparity between first order methods and second order methods by continuing to use gradient information to calculate Hessian approximations. We provide empirical results on the classification task of the MNIST dataset and show robust convergence with preferred generalization characteristics.

Generating adversarial samples is gathering much attention as an intuitive approach to evaluate the robustness of learning models. Extensive recent works have demonstrated that numerous advanced image classifiers are defenseless to adversarial perturbations in the white-box setting. However, the white-box setting assumes attackers to have prior knowledge of model parameters, which are generally inaccessible in real world cases. In this paper, we concentrate on the hard-label black-box setting where attackers can only pose queries to probe the model parameters responsible for classifying different images. Therefore, the issue is converted into minimizing non-continuous function. A black-box approach is proposed to address both massive queries and the non-continuous step function problem by applying a combination of a linear fine-grained search, Fibonacci search, and a zeroth order optimization algorithm. However, the input dimension of a image is so high that the estimation of gradient is noisy. Hence, we adopt a zeroth-order optimization method in high dimensions. The approach converts calculation of gradient into a linear regression model and extracts dimensions that are more significant. Experimental results illustrate that our approach can relatively reduce the amount of queries and effectively accelerate convergence of the optimization method.

As automatic speech recognition (ASR) systems have been integrated into a diverse set of devices around us in recent years, security vulnerabilities of them have become an increasing concern for the public. Existing studies have demonstrated that deep neural networks (DNNs), acting as the computation core of ASR systems, is vulnerable to deliberately designed adversarial attacks. Based on the gradient descent algorithm, existing studies have successfully generated adversarial samples which can disturb ASR systems and produce adversary-expected transcript texts designed by adversaries. Most of these research simulated white-box attacks which require knowledge of all the components in the targeted ASR systems. In this work, we propose the first semi-black-box attack against the ASR system - Kaldi. Requiring only partial information from Kaldi and none from DNN, we can embed malicious commands into a single audio chip based on the gradient-independent genetic algorithm. The crafted audio clip could be recognized as the embedded malicious commands by Kaldi and unnoticeable to humans in the meanwhile. Experiments show that our attack can achieve high attack success rate with unnoticeable perturbations to three types of audio clips (pop music, pure music, and human command) without the need of the underlying DNN model parameters and architecture.

The application of deep recurrent networks to audio transcription has led to impressive gains in automatic speech recognition (ASR) systems. Many have demonstrated that small adversarial perturbations can fool deep neural networks into incorrectly predicting a specified target with high confidence. Current work on fooling ASR systems have focused on white-box attacks, in which the model architecture and parameters are known. In this paper, we adopt a black-box approach to adversarial generation, combining the approaches of both genetic algorithms and gradient estimation to solve the task. We achieve a 89.25% targeted attack similarity, with 35% targeted attack success rate, after 3000 generations while maintaining 94.6% audio file similarity.

The storage efficiency of hash codes and their application in the fast approximate nearest neighbor search, along with the explosion in the size of available labeled image datasets caused an intensive interest in developing learning based hash algorithms recently. In this paper, we present a learning based hash algorithm that utilize ordinal information of feature vectors. We have proposed a novel mathematically differentiable approximation of argmax function for this hash algorithm. It has enabled seamless integration of hash function with deep neural network architecture which can exploit the rich feature vectors generated by convolutional neural networks. We have also proposed a loss function for the case that the hash code is not binary and its entries are digits of arbitrary k-ary base. The resultant model comprised of feature vector generation and hashing layer is amenable to end-to-end training using gradient descent methods. In contrast to the majority of current hashing algorithms that are either not learning based or use hand-crafted feature vectors as input, simultaneous training of the components of our system results in better optimization. Extensive evaluations on NUS-WIDE, CIFAR-10 and MIRFlickr benchmarks show that the proposed algorithm outperforms state-of-art and classical data agnostic, unsupervised and supervised hashing methods by 2.6% to 19.8% mean average precision under various settings.

Aiming at the problem of node identification in wireless networks, a method of node identification based on deep learning is proposed, which starts with the tiny features of nodes in radiofrequency layer. Firstly, in order to cut down the computational complexity, Principal Component Analysis is used to reduce the dimension of node sample data. Secondly, a convolution neural network containing two hidden layers is designed to extract local features of the preprocessed data. Stochastic gradient descent method is used to optimize the parameters, and the Softmax Model is used to determine the output label. Finally, the effectiveness of the method is verified by experiments on practical wireless ad-hoc network.

Deep neural networks are susceptible to various inference attacks as they remember information about their training data. We design white-box inference attacks to perform a comprehensive privacy analysis of deep learning models. We measure the privacy leakage through parameters of fully trained models as well as the parameter updates of models during training. We design inference algorithms for both centralized and federated learning, with respect to passive and active inference attackers, and assuming different adversary prior knowledge. We evaluate our novel white-box membership inference attacks against deep learning algorithms to trace their training data records. We show that a straightforward extension of the known black-box attacks to the white-box setting (through analyzing the outputs of activation functions) is ineffective. We therefore design new algorithms tailored to the white-box setting by exploiting the privacy vulnerabilities of the stochastic gradient descent algorithm, which is the algorithm used to train deep neural networks. We investigate the reasons why deep learning models may leak information about their training data. We then show that even well-generalized models are significantly susceptible to white-box membership inference attacks, by analyzing state-of-the-art pre-trained and publicly available models for the CIFAR dataset. We also show how adversarial participants, in the federated learning setting, can successfully run active membership inference attacks against other participants, even when the global model achieves high prediction accuracies.

The enormous growth of Internet-based traffic exposes corporate networks with a wide variety of vulnerabilities. Intrusive traffics are affecting the normal functionality of network's operation by consuming corporate resources and time. Efficient ways of identifying, protecting, and mitigating from intrusive incidents enhance productivity. As Intrusion Detection System (IDS) is hosted in the network and at the user machine level to oversee the malicious traffic in the network and at the individual computer, it is one of the critical components of a network and host security. Unsupervised anomaly traffic detection techniques are improving over time. This research aims to find an efficient classifier that detects anomaly traffic from NSL-KDD dataset with high accuracy level and minimal error rate by experimenting with five machine learning techniques. Five binary classifiers: Stochastic Gradient Decent, Random Forests, Logistic Regression, Support Vector Machine, and Sequential Model are tested and validated to produce the result. The outcome demonstrates that Random Forest Classifier outperforms the other four classifiers with and without applying the normalization process to the dataset.

From signal processing to emerging deep neural networks, a range of applications exhibit intrinsic error resilience. For such applications, approximate computing opens up new possibilities for energy-efficient computing by producing slightly inaccurate results using greatly simplified hardware. Adopting this approach, a variety of basic arithmetic units, such as adders and multipliers, have been effectively redesigned to generate approximate results for many error-resilient applications.In this work, we propose SECO, an approximate exponential function unit (EFU). Exponentiation is a key operation in many signal processing applications and more importantly in spiking neuron models, but its energy-efficient implementation has been inadequately explored. We also introduce a cross-layer design method for SECO to optimize the energy-accuracy trade-off. At the algorithm level, SECO offers runtime scaling between energy efficiency and accuracy based on approximate Taylor expansion, where the error is minimized by optimizing parameters using discrete gradient descent at design time. At the circuit level, our error analysis method efficiently explores the design space to select the energy-accuracy-optimal approximate multiplier at design time. In tandem, the cross-layer design and runtime optimization method are able to generate energy-efficient and accurate approximate EFU designs that are up to 99.7% accurate at a power consumption of 3.73 pJ per exponential operation. SECO is also evaluated on the adaptive exponential integrate-and-fire neuron model, yielding only 0.002% timing error and 0.067% value error compared to the precise neuron model.