Biblio

Although sequence-to-sequence attentional neural machine translation (NMT) has achieved great progress recently, it is confronted with two challenges: learning optimal model parameters for long parallel sentences and well exploiting different scopes of contexts. In this paper, partially inspired by the idea of segmenting a long sentence into short clauses, each of which can be easily translated by NMT, we propose a hierarchy-to-sequence attentional NMT model to handle these two challenges. Our encoder takes the segmented clause sequence as input and explores a hierarchical neural network structure to model words, clauses, and sentences at different levels, particularly with two layers of recurrent neural networks modeling semantic compositionality at the word and clause level. Correspondingly, the decoder sequentially translates segmented clauses and simultaneously applies two types of attention models to capture contexts of interclause and intraclause for translation prediction. In this way, we can not only improve parameter learning, but also well explore different scopes of contexts for translation. Experimental results on Chinese-English and English-German translation demonstrate the superiorities of the proposed model over the conventional NMT model.

In this paper, we consider the problem of decentralized verification for large-scale cascade interconnections of linear subsystems such that dissipativity properties of the overall system are guaranteed with minimum knowledge of the dynamics. In order to achieve compositionality, we distribute the verification process among the individual subsystems, which utilize limited information received locally from their immediate neighbors. Furthermore, to obviate the need for full knowledge of the subsystem parameters, each decentralized verification rule employs a model-free learning structure; a reinforcement learning algorithm that allows for online evaluation of the appropriate storage function that can be used to verify dissipativity of the system up to that point. Finally, we show how the interconnection can be extended by adding learning-enabled subsystems while ensuring dissipativity.

While recent studies have shed light on the expressivity, complexity and compositionality of convolutional networks, the real inductive bias of the family of functions reachable by gradient descent on natural data is still unknown. By exploiting symmetries in the preactivation space of convolutional layers, we present preliminary empirical evidence of regularities in the preimage of trained rectifier networks, in terms of arrangements of polytopes, and relate it to the nonlinear transformations applied by the network to its input.

Neural architectures are the foundation for improving performance of deep neural networks (DNNs). This paper presents deep compositional grammatical architectures which harness the best of two worlds: grammar models and DNNs. The proposed architectures integrate compositionality and reconfigurability of the former and the capability of learning rich features of the latter in a principled way. We utilize AND-OR Grammar (AOG) as network generator in this paper and call the resulting networks AOGNets. An AOGNet consists of a number of stages each of which is composed of a number of AOG building blocks. An AOG building block splits its input feature map into N groups along feature channels and then treat it as a sentence of N words. It then jointly realizes a phrase structure grammar and a dependency grammar in bottom-up parsing the “sentence” for better feature exploration and reuse. It provides a unified framework for the best practices developed in state-of-the-art DNNs. In experiments, AOGNet is tested in the ImageNet-1K classification benchmark and the MS-COCO object detection and segmentation benchmark. In ImageNet-1K, AOGNet obtains better performance than ResNet and most of its variants, ResNeXt and its attention based variants such as SENet, DenseNet and DualPathNet. AOGNet also obtains the best model interpretability score using network dissection. AOGNet further shows better potential in adversarial defense. In MS-COCO, AOGNet obtains better performance than the ResNet and ResNeXt backbones in Mask R-CNN.

Recent advances in machine learning enable wider applications of prediction models in cyber-physical systems. Smart grids are increasingly using distributed sensor settings for distributed sensor fusion and information processing. Load forecasting systems use these sensors to predict future loads to incorporate into dynamic pricing of power and grid maintenance. However, these inference predictors are highly complex and thus vulnerable to adversarial attacks. Moreover, the adversarial attacks are synthetic norm-bounded modifications to a limited number of sensors that can greatly affect the accuracy of the overall predictor. It can be much cheaper and effective to incorporate elements of security and resilience at the earliest stages of design. In this paper, we demonstrate how to analyze the security and resilience of learning-based prediction models in power distribution networks by utilizing a domain-specific deep-learning and testing framework. This framework is developed using DeepForge and enables rapid design and analysis of attack scenarios against distributed smart meters in a power distribution network. It runs the attack simulations in the cloud backend. In addition to the predictor model, we have integrated an anomaly detector to detect adversarial attacks targeting the predictor. We formulate the stealthy adversarial attacks as an optimization problem to maximize prediction loss while minimizing the required perturbations. Under the worst-case setting, where the attacker has full knowledge of both the predictor and the detector, an iterative attack method has been developed to solve for the adversarial perturbation. We demonstrate the framework capabilities using a GridLAB-D based power distribution network model and show how stealthy adversarial attacks can affect smart grid prediction systems even with a partial control of network.

High frequency of network security incidents has also brought a lot of negative effects and even huge economic losses to countries, enterprises and individuals in recent years. Therefore, more and more attention has been paid to the problem of network security. In order to evaluate the newly included vulnerability text information accurately, and to reduce the workload of experts and the false negative rate of the traditional method. Multiple deep learning methods for vulnerability text classification evaluation are proposed in this paper. The standard Cross Site Scripting (XSS) vulnerability text data is processed first, and then classified using three kinds of deep neural networks (CNN, LSTM, TextRCNN) and one kind of traditional machine learning method (XGBoost). The dropout ratio of the optimal CNN network, the epoch of all deep neural networks and training set data were tuned via experiments to improve the fit on our target task. The results show that the deep learning methods evaluate vulnerability risk levels better, compared with traditional machine learning methods, but cost more time. We train our models in various training sets and test with the same testing set. The performance and utility of recurrent convolutional neural networks (TextRCNN) is highest in comparison to all other methods, which classification accuracy rate is 93.95%.

In the Web 2.0 era, user interaction makes Web application more diverse, but brings threats, among which XSS vulnerability is the common and pernicious one. In order to promote the efficiency of XSS detection, this paper investigates the parameter characteristics of malicious XSS attacks. We identify whether a parameter is malicious or not through detecting user input parameters with SVM algorithm. The original malicious XSS parameters are deformed by DQN algorithm for reinforcement learning for rule-based WAF to be anti-anti-virus. Based on this method, we can identify whether a specific WAF is secure. The above model creates a more efficient automatic XSS detection tool and a more targeted automatic anti-anti-virus payload generation tool. This paper also explores the automatic generation of XSS attack codes with RNN LSTM algorithm.

Cross site scripting (XSS) attack is one of the attacks on the web. It brings session hijack with HTTP cookies, information collection with fake HTML input form and phishing with dummy sites. As a countermeasure of XSS attack, machine learning has attracted a lot of attention. There are existing researches in which SVM, Random Forest and SCW are used for the detection of the attack. However, in the researches, there are problems that the size of data set is too small or unbalanced, and that preprocessing method for vectorization of strings causes misclassification. The highest accuracy of the classification was 98% in existing researches. Therefore, in this paper, we improved the preprocessing method for vectorization by using word2vec to find the frequency of appearance and co-occurrence of the words in XSS attack scripts. Moreover, we also used a large data set to decrease the deviation of the data. Furthermore, we evaluated the classification results with two procedures. One is an inappropriate procedure which some researchers tend to select by mistake. The other is an appropriate procedure which can be applied to an attack detection filter in the real environment.

Cyber-physical systems (CPS) consist of sensors, actuators, and controllers all communicating over a network; if any subset becomes compromised, an attacker could cause significant damage. With access to data logs and a model of the CPS, the physical effects of an attack could potentially be detected before any damage is done. Manually building a model that is accurate enough in practice, however, is extremely difficult. In this paper, we propose a novel approach for constructing models of CPS automatically, by applying supervised machine learning to data traces obtained after systematically seeding their software components with faults ("mutants"). We demonstrate the efficacy of this approach on the simulator of a real-world water purification plant, presenting a framework that automatically generates mutants, collects data traces, and learns an SVM-based model. Using cross-validation and statistical model checking, we show that the learnt model characterises an invariant physical property of the system. Furthermore, we demonstrate the usefulness of the invariant by subjecting the system to 55 network and code-modification attacks, and showing that it can detect 85% of them from the data logs generated at runtime.

The measurement and provision of precise and up-to-date traffic-related key performance indicators is a key element and crucial factor for intelligent traffic control systems in upcoming smart cities. The street network is considered as a highly-dynamic Cyber Physical System (CPS) where measured information forms the foundation for dynamic control methods aiming to optimize the overall system state. Apart from global system parameters like traffic flow and density, specific data, such as velocity of individual vehicles as well as vehicle type information, can be leveraged for highly sophisticated traffic control methods like dynamic type-specific lane assignments. Consequently, solutions for acquiring these kinds of information are required and have to comply with strict requirements ranging from accuracy over cost-efficiency to privacy preservation. In this paper, we present a system for classifying vehicles based on their radio-fingerprint. In contrast to other approaches, the proposed system is able to provide real-time capable and precise vehicle classification as well as cost-efficient installation and maintenance, privacy preservation and weather independence. The system performance in terms of accuracy and resource-efficiency is evaluated in the field using comprehensive measurements. Using a machine learning based approach, the resulting success ratio for classifying cars and trucks is above 99%.

In recent years, sentiment analysis has been aware within software engineering domain. While automated sentiment analysis has long been suffering from doubt of accuracy, the tool performance is unstable when being applied on datasets other than the original dataset for evaluation. Researchers also have the disagreements upon if machine learning algorithms perform better than conventional lexicon and rule based approaches. In this paper, we looked into the factors in datasets that may affect the evaluation performance, also evaluated the popular machine learning algorithms in sentiment analysis, then proposed a novel structure for automated sentiment tool combines advantages from both approaches.

To detect specific types of bugs and vulnerabilities, static analysis tools must be correctly configured with security-relevant methods (SRM), e.g., sources, sinks, sanitizers and authentication methods-usually a very labour-intensive and error-prone process. This work presents the semi-automated tool SWAN\_ASSIST, which aids the configuration with an IntelliJ plugin based on active machine learning. It integrates our novel automated machine-learning approach SWAN, which identifies and classifies Java SRM. SWAN\_ASSIST further integrates user feedback through iterative learning. SWAN\_ASSIST aids developers by asking them to classify at each point in time exactly those methods whose classification best impact the classification result. Our experiments show that SWAN\_ASSIST classifies SRM with a high precision, and requires a relatively low effort from the user. A video demo of SWAN\_ASSIST can be found at https://youtu.be/fSyD3V6EQOY. The source code is available at https://github.com/secure-software-engineering/swan.

We study a model where a data collector obtains data from users through a payment mechanism to learn the underlying state from the elicited data. The private signal of each user represents her individual knowledge about the state. Through social interactions, each user can also learn noisy versions of her friends' signals, which is called group signals. Based on both her private signal and group signals, each user makes strategic decisions to report a privacy-preserved version of her data to the data collector. We develop a Bayesian game theoretic framework to study the impact of social learning on users' data reporting strategies and devise the payment mechanism for the data collector accordingly. Our findings reveal that, the Bayesian-Nash equilibrium can be in the form of either a symmetric randomized response (SR) strategy or an informative non-disclosive (ND) strategy. A generalized majority voting rule is applied by each user to her noisy group signals to determine which strategy to follow. When a user plays the ND strategy, she reports privacy-preserving data completely based on her group signals, independent of her private signal, which indicates that her privacy cost is zero. Both the data collector and the users can benefit from social learning which drives down the privacy costs and helps to improve the state estimation at a given payment budget. We derive bounds on the minimum total payment required to achieve a given level of state estimation accuracy.

With the popularity of smart devices and the widespread use of the Wi-Fi-based indoor localization, edge computing is becoming the mainstream paradigm of processing massive sensing data to acquire indoor localization service. However, these data which were conveyed to train the localization model unintentionally contain some sensitive information of users/devices, and were released without any protection may cause serious privacy leakage. To solve this issue, we propose a lightweight differential privacy-preserving mechanism for the edge computing environment. We extend ε-differential privacy theory to a mature machine learning localization technology to achieve privacy protection while training the localization model. Experimental results on multiple real-world datasets show that, compared with the original localization technology without privacy-preserving, our proposed scheme can achieve high accuracy of indoor localization while providing differential privacy guarantee. Through regulating the value of ε, the data quality loss of our method can be controlled up to 8.9% and the time consumption can be almost negligible. Therefore, our scheme can be efficiently applied in the edge networks and provides some guidance on indoor localization privacy protection in the edge computing.

Deep neural networks (DNNs) have demonstrated impressive performance on many challenging machine learning tasks. However, DNNs are vulnerable to adversarial inputs generated by adding maliciously crafted perturbations to the benign inputs. As a growing number of attacks have been reported to generate adversarial inputs of varying sophistication, the defense-attack arms race has been accelerated. In this paper, we present MODEF, a cross-layer model diversity ensemble framework. MODEF intelligently combines unsupervised model denoising ensemble with supervised model verification ensemble by quantifying model diversity, aiming to boost the robustness of the target model against adversarial examples. Evaluated using eleven representative attacks on popular benchmark datasets, we show that MODEF achieves remarkable defense success rates, compared with existing defense methods, and provides a superior capability of repairing adversarial inputs and making correct predictions with high accuracy in the presence of black-box attacks.

It is critical to secure the Industrial Internet of Things (IIoT) devices because of potentially devastating consequences in case of an attack. Machine learning (ML) and big data analytics are the two powerful leverages for analyzing and securing the Internet of Things (IoT) technology. By extension, these techniques can help improve the security of the IIoT systems as well. In this paper, we first present common IIoT protocols and their associated vulnerabilities. Then, we run a cyber-vulnerability assessment and discuss the utilization of ML in countering these susceptibilities. Following that, a literature review of the available intrusion detection solutions using ML models is presented. Finally, we discuss our case study, which includes details of a real-world testbed that we have built to conduct cyber-attacks and to design an intrusion detection system (IDS). We deploy backdoor, command injection, and Structured Query Language (SQL) injection attacks against the system and demonstrate how a ML-based anomaly detection system can perform well in detecting these attacks. We have evaluated the performance through representative metrics to have a fair point of view on the effectiveness of the methods.

This paper investigates the effectiveness of reinforcement learning (RL) model in clustering as an approach to achieve higher network scalability in distributed cognitive radio networks. Specifically, it analyzes the effects of RL parameters, namely the learning rate and discount factor in a volatile environment, which consists of member nodes (or secondary users) that launch attacks with various probabilities of attack. The clusterhead, which resides in an operating region (environment) that is characterized by the probability of attacks, countermeasures the malicious SUs by leveraging on a RL model. Simulation results have shown that in a volatile operating environment, the RL model with learning rate α= 1 provides the highest network scalability when the probability of attacks ranges between 0.3 and 0.7, while the discount factor γ does not play a significant role in learning in an operating environment that is volatile due to attacks.

In the deep learning tasks, we can design different network models to address different tasks (classification, detection, segmentation). But traditional deep learning networks simply increase the depth and breadth of the network. This leads to a higher complexity of the model. We propose Adaptively Weighted Channel Feature Network of Mixed Convolution Kernel(SKENet). SKENet extract features from different kernels, then mixed those features by elementwise, lastly do sigmoid operator on channel features to get adaptive weightings. We did a simple classification test on the CIFAR10 amd CIFAR100 dataset. The results show that SKENet can achieve a better result in a shorter time. After that, we did an object detection experiment on the VOC dataset. The experimental results show that SKENet is far ahead of the SKNet[20] in terms of speed and accuracy.

To solve the problems of weak linkage ability and low intellectualization of strategy allocation in existing network security devices, a new method of cognitive linkage of network security equipment is proposed by learning from human brain. Firstly, the basic connotation and cognitive cycle of cognitive linkage are expounded. Secondly, the main functions of cognitive linkage are clarified. Finally, the cognitive linkage system model is constructed, and the information process flow of cognitive linkage is described. Cognitive linkage of network security equipment provides a new way to effectively enhance the overall protection capability of network security equipment.

The number of phishing attacks has increased in Latin America, exceeding the operational skills of cybersecurity analysts. The cognitive security application proposes the use of bigdata, machine learning, and data analytics to improve response times in attack detection. This paper presents an investigation about the analysis of anomalous behavior related with phishing web attacks and how machine learning techniques can be an option to face the problem. This analysis is made with the use of an contaminated data sets, and python tools for developing machine learning for detect phishing attacks through of the analysis of URLs to determinate if are good or bad URLs in base of specific characteristics of the URLs, with the goal of provide realtime information for take proactive decisions that minimize the impact of an attack.

We propose a novel hybrid imaging system to acquire 4D high-speed hyperspectral (HSHS) videos with high spatial and spectral resolution. The proposed system consists of two branches: one branch performs Nyquist sampling in the temporal dimension while integrating the whole spectrum, resulting in a high-frame-rate panchromatic video; the other branch performs compressive sampling in the spectral dimension with longer exposures, resulting in a low-frame-rate hyperspectral video. Owing to the high light throughput and complementary sampling, these two branches jointly provide reliable measurements for recovering the underlying HSHS video. Moreover, the panchromatic video can be used to learn an over-complete 3D dictionary to represent each band-wise video sparsely, thanks to the inherent structural similarity in the spectral dimension. Based on the joint measurements and the self-adaptive dictionary, we further propose a simultaneous spectral sparse (3S) model to reinforce the structural similarity across different bands and develop an efficient computational reconstruction algorithm to recover the HSHS video. Both simulation and hardware experiments validate the effectiveness of the proposed approach. To the best of our knowledge, this is the first time that hyperspectral videos can be acquired at a frame rate up to 100fps with commodity optical elements and under ordinary indoor illumination.

Vehicular ad hoc networks (VANETs) are designed to provide traffic safety by enabling vehicles to broadcast information-such as speed, location and heading-through inter-vehicular communications to proactively avoid collisions. However, the attacks targeting these networks might overshadow their advantages if not protected against. One powerful threat against VANETs is vehicular botnets. In our earlier work, we demonstrated several vehicular botnet attacks that can have damaging impacts on the security and privacy of VANETs. In this paper, we present SHIELDNET, the first detection mechanism against vehicular botnets. Similar to the detection approaches against Internet botnets, we target the vehicular botnet communication and use several machine learning techniques to identify vehicular bots. We show via simulation that SHIELDNET can identify 77 percent of the vehicular bots. We propose several improvements on the VANET standards and show that their existing vulnerabilities make an effective defense against vehicular botnets infeasible.

The accuracy of the image classification by the convolutional neural network is exceeding the ability of human being and contributes to various fields. However, the improvement of the image recognition technology gives a great blow to security system with an image such as CAPTCHA. In particular, since the character string CAPTCHA has already added distortion and noise in order not to be read by the computer, it becomes a problem that the human readability is lowered. Adversarial examples is a technique to produce an image letting an image classification by the machine learning be wrong intentionally. The best feature of this technique is that when human beings compare the original image with the adversarial examples, they cannot understand the difference on appearance. However, Adversarial examples that is created with conventional FGSM cannot completely misclassify strong nonlinear networks like CNN. Osadchy et al. have researched to apply this adversarial examples to CAPTCHA and attempted to let CNN misclassify them. However, they could not let CNN misclassify character images. In this research, we propose a method to apply FGSM to the character string CAPTCHAs and to let CNN misclassified them.

CAPTCHA is a web-based authentication method used by websites to distinguish between humans (valid users) and bots (attackers). Audio captcha is an accessible captcha meant for the visually disabled section of users such as color-blind, blind, near-sighted users. Firstly, this paper analyzes how secure current audio captchas are from attacks using machine learning (ML) and deep learning (DL) models. Each audio captcha is made up of five, seven or ten random digits[0-9] spoken one after the other along with varying background noise throughout the length of the audio. If the ML or DL model is able to correctly identify all spoken digits and in the correct order of occurance in a single audio captcha, we consider that captcha to be broken and the attack to be successful. Throughout the paper, accuracy refers to the attack model's success at breaking audio captchas. The higher the attack accuracy, the more unsecure the audio captchas are. In our baseline experiments, we found that attack models could break audio captchas that had no background noise or medium background noise with any number of spoken digits with nearly 99% to 100% accuracy. Whereas, audio captchas with high background noise were relatively more secure with attack accuracy of 85%. Secondly, we propose that the concepts of adversarial examples algorithms can be used to create a new kind of audio captcha that is more resilient towards attacks. We found that even after retraining the models on the new adversarial audio data, the attack accuracy remained as low as 25% to 36% only. Lastly, we explore the benefits of creating adversarial audio captcha through different algorithms such as Basic Iterative Method (BIM) and deepFool. We found that as long as the attacker has less than 45% sample from each kinds of adversarial audio datasets, the defense will be successful at preventing attacks.

A detailed model of an attack on the power grid involves both a preparation stage as well as an execution stage of the attack. This paper introduces a novel Hybrid Attack Model (HAM) that combines Probabilistic Learning Attacker, Dynamic Defender (PLADD) model and a Markov Chain model to simulate the planning and execution stages of a bad data injection attack in power grid. We discuss the advantages and limitations of the prior work models and of our proposed Hybrid Attack Model and show that HAM is more effective compared to individual PLADD or Markov Chain models.