Biblio

Negotiation is a complex social interaction that encapsulates emotional encounters in human decision-making. Virtual agents that can negotiate with humans are useful in pedagogy and conversational AI. To advance the development of such agents, we explore the prediction of two important subjective goals in a negotiation – outcome satisfaction and partner perception. Specifically, we analyze the extent to which emotion attributes extracted from the negotiation help in the prediction, above and beyond the individual difference variables. We focus on a recent dataset in chat-based negotiations, grounded in a realistic camping scenario. We study three degrees of emotion dimensions – emoticons, lexical, and contextual by leveraging affective lexicons and a state-of-the-art deep learning architecture. Our insights will be helpful in designing adaptive negotiation agents that interact through realistic communication interfaces.

Conversational agents are the most popular AI technology in IT trends. Domain specific chatbots are now used by almost every industry in order to upgrade their customer service. The Proposed paper shows the modelling and performance of one such conversational agent created using deep learning. The proposed model utilizes NMT (Neural Machine Translation) from the TensorFlow software libraries. A BiRNN (Bidirectional Recurrent Neural Network) is used in order to process input sentences that contain large number of tokens (20-40 words). In order to understand the context of the input sentence attention model is used along with BiRNN. The conversational models usually have one drawback, that is, they sometimes provide irrelevant answer to the input. This happens quite often in conversational chatbots as the chatbot doesn't realize that it is answering without context. This drawback is solved in the proposed system using Deep Reinforcement Learning technique. Deep reinforcement Learning follows a reward system that enables the bot to differentiate between right and wrong answers. Deep Reinforcement Learning techniques allows the chatbot to understand the sentiment of the query and reply accordingly. The Deep Reinforcement Learning algorithms used in the proposed system is Q-Learning, Deep Q Neural Network (DQN) and Distributional Reinforcement Learning with Quantile Regression (QR-DQN). The performance of each algorithm is evaluated and compared in this paper in order to find the best DRL algorithm. The dataset used in the proposed system is Cornell Movie-dialogs corpus and CoQA (A Conversational Question Answering Challenge). CoQA is a large dataset that contains data collected from 8000+ conversations in the form of questions and answers. The main goal of the proposed work is to increase the relevancy of the chatbot responses and to increase the perplexity of the conversational chatbot.

Empathy is a fundamental mechanism of human interactions. As such, it should be an integral part of Human-Computer Interaction systems to make them more relatable. With this work, we focused on conversational scenarios where integrating empathy is crucial to perceive the computer like a human. As a result, we derived the high-level architecture of an Empathetic Conversational Agent we are willing to implement. We relied on theories about artificial empathy to derive the function approximating this mechanism and selected the conversational aspects to control for an empathetic interaction. In particular, we designed a core empathetic controller manages the empathetic responses, predicting, at each turn, the high-level content of the response. The derived architecture integrates empathy in a task-agnostic manner; hence we can employ it in multiple scenarios by changing the objective of the controller.

Visual Question Answering (VQA) models have achieved significant success in recent times. Despite the success of VQA models, they are mostly black-box models providing no reasoning about the predicted answer, thus raising questions for their applicability in safety-critical such as autonomous systems and cyber-security. Current state of the art fail to better complex questions and thus are unable to exploit compositionality. To minimize the black-box effect of these models and also to make them better exploit compositionality, we propose a Dynamic Neural Network (DMN), which can understand a particular question and then dynamically assemble various relatively shallow deep learning modules from a pool of modules to form a network. We incorporate compositional temporal attention to these deep learning based modules to increase compositionality exploitation. This results in achieving better understanding of complex questions and also provides reasoning as to why the module predicts a particular answer. Experimental analysis on the two benchmark datasets, VQA2.0 and CLEVR, depicts that our model outperforms the previous approaches for Visual Question Answering task as well as provides better reasoning, thus making it reliable for mission critical applications like safety and security.

In this conceptual work, we present Deep Convolutional Gaussian Mixture Models (DCGMMs): a new formulation of deep hierarchical Gaussian Mixture Models (GMMs) that is particularly suitable for describing and generating images. Vanilla (i.e., flat) GMMs require a very large number of components to describe images well, leading to long training times and memory issues. DCGMMs avoid this by a stacked architecture of multiple GMM layers, linked by convolution and pooling operations. This allows to exploit the compositionality of images in a similar way as deep CNNs do. DCGMMs can be trained end-to-end by Stochastic Gradient Descent. This sets them apart from vanilla GMMs which are trained by Expectation-Maximization, requiring a prior k-means initialization which is infeasible in a layered structure. For generating sharp images with DCGMMs, we introduce a new gradient-based technique for sampling through non-invertible operations like convolution and pooling. Based on the MNIST and FashionMNIST datasets, we validate the DCGMMs model by demonstrating its superiority over flat GMMs for clustering, sampling and outlier detection.

The software industry is dominating many are like health care, finance, agriculture and entertainment. Software security has become an essential issue-outsider libraries, which assume a significant part in programming. The finding weaknesses in the binary code is a significant issue that presently cannot seem to be handled, as showed by numerous weaknesses wrote about an everyday schedule. Software seller sells the software to the client if the client wants to check the software's vulnerability it is a cumbersome task. Presently many deep learning-based methods also introduced to find the security weakness in the binary code. This paper present the merits and demerits of binary code analysis used by a different method.

In order to improve the security and reliability of information system and reduce the risk of vulnerability intrusion and attack, an automatic patch installation method of operating systems based on deep learning is proposed, If the installation is successful, the basic information of the system will be returned to the visualization server. If the installation fails, it is recommended to upgrading manually and display it on the patch detection visualization server. Through the practical application of statistical analysis, the statistical results show that the proposed method is significantly better than the original and traditional installation methods, which can effectively avoid the problem of client repeated download, and greatly improve the success rate of patch automatic upgrades. It effectively saves the upgrade cost and ensures the security and reliability of the information system.

As a large amount of renewable energy is injected into the power grid, the source side of the power grid becomes extremely uncertain. Traditional static safety analysis methods based on pure physical models can no longer quickly and reliably give analysis results. Therefore, this paper proposes a deep learning-based static security analytical method. First, the static security assessment index of the power grid under the N-1 principle is proposed. Secondly, a neural network model and its input and output data for static safety analysis problems are designed. Finally, the validity of the proposed method was verified by IEEE grid data. Experiments show that the proposed method can quickly and accurately give the static security analysis results of the source-side high uncertainty grid.

With the development of 5G technology and intelligent terminals, the future direction of the Industrial Internet of Things (IIoT) evolution is Pervasive Edge Computing (PEC). In the pervasive edge computing environment, intelligent terminals can perform calculations and data processing. By migrating part of the original cloud computing model's calculations to intelligent terminals, the intelligent terminal can complete model training without uploading local data to a remote server. Pervasive edge computing solves the problem of data islands and is also successfully applied in scenarios such as vehicle interconnection and video surveillance. However, pervasive edge computing is facing great security problems. Suppose the remote server is honest but curious. In that case, it can still design algorithms for the intelligent terminal to execute and infer sensitive content such as their identity data and private pictures through the information returned by the intelligent terminal. In this paper, we research the problem of honest but curious remote servers infringing intelligent terminal privacy and propose a differential privacy collaborative deep learning algorithm in the pervasive edge computing environment. We use a Gaussian mechanism that meets the differential privacy guarantee to add noise on the first layer of the neural network to protect the data of the intelligent terminal and use analytical moments accountant technology to track the cumulative privacy loss. Experiments show that with the Gaussian mechanism, the training data of intelligent terminals can be protected reduction inaccuracy.

Recommender systems, which aim to suggest personalized lists of items for users, have drawn a lot of attention. In fact, many of these state-of-the-art recommender systems have been built on deep neural networks (DNNs). Recent studies have shown that these deep neural networks are vulnerable to attacks, such as data poisoning, which generate fake users to promote a selected set of items. Correspondingly, effective defense strategies have been developed to detect these generated users with fake profiles. Thus, new strategies of creating more ‘realistic’ user profiles to promote a set of items should be investigated to further understand the vulnerability of DNNs based recommender systems. In this work, we present a novel framework CopyAttack. It is a reinforcement learning based black-box attacking method that harnesses real users from a source domain by copying their profiles into the target domain with the goal of promoting a subset of items. CopyAttack is constructed to both efficiently and effectively learn policy gradient networks that first select, then further refine/craft user profiles from the source domain, and ultimately copy them into the target domain. CopyAttack’s goal is to maximize the hit ratio of the targeted items in the Top-k recommendation list of the users in the target domain. We conducted experiments on two real-world datasets and empirically verified the effectiveness of the proposed framework. The implementation of CopyAttack is available at https://github.com/wenqifan03/CopyAttack.

In Social Aware Network (SAN) model, the elementary actions focus on investigating the attributes and behaviors of the customer. This analysis of customer attributes facilitate in the design of highly active and improved protocols. In specific, the recommender systems are highly vulnerable to the shilling attack. The recommender system provides the solution to solve the issues like information overload. Collaborative filtering based recommender systems are susceptible to shilling attack known as profile injection attacks. In the shilling attack, the malicious users bias the output of the system's recommendations by adding the fake profiles. The attacker exploits the customer reviews, customer ratings and fake data for the processing of recommendation level. It is essential to detect the shilling attack in the network for sustaining the reliability and fairness of the recommender systems. This article reviews the most prominent issues and challenges of shilling attack. This paper presents the literature survey which is contributed in focusing of shilling attack and also describes the merits and demerits with its evaluation metrics like attack detection accuracy, precision and recall along with different datasets used for identifying the shilling attack in SAN network.

Modern smart grid systems are heavily dependent on Information and Communication Technology, and this dependency makes them prone to cyber-attacks. The occurrence of a cyber-attack has increased in recent years resulting in substantial damage to power systems. For a reliable and stable operation, cyber protection, control, and detection techniques are becoming essential. Automated detection of cyberattacks with high accuracy is a challenge. To address this, we propose a two-layer hierarchical machine learning model having an accuracy of 95.44 % to improve the detection of cyberattacks. The first layer of the model is used to distinguish between the two modes of operation - normal state or cyberattack. The second layer is used to classify the state into different types of cyberattacks. The layered approach provides an opportunity for the model to focus its training on the targeted task of the layer, resulting in improvement in model accuracy. To validate the effectiveness of the proposed model, we compared its performance against other recent cyber attack detection models proposed in the literature.

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.

In this study, it is aimed to reduce False-Alarm levels and increase the correct detection rate in order to reduce this uncertainty. Within the scope of the study, 13157 SQLi and XSS type malicious and 10000 normal HTTP Requests were used. All HTTP requests were received from the same web server, and it was observed that normal requests and malicious requests were close to each other. In this study, a novel approach is presented via both digitization and expressing the data with words in the data preprocessing stages. LSTM, MLP, CNN, GNB, SVM, KNN, DT, RF algorithms were used for classification and the results were evaluated with accuracy, precision, recall and F1-score metrics. As a contribution of this study, we can clearly express the following inferences. Each payload even if it seems different which has the same impact maybe that we can clearly view after the preprocessing phase. After preprocessing we are calculating euclidean distances which brings and gives us the relativity between expressions. When we put this relativity as an entry data to machine learning and/or deep learning models, perhaps we can understand the benign request or the attack vector difference.

While various encryption algorithms ensure data security, it is essential to determine the accuracy and loss values and performance status in the analyzes made to determine encrypted data by deep learning. In this research, the analysis steps made by applying deep learning methods to encrypted cifar10 picture data are presented practically. The data was tried to be estimated by training with VGG16, VGG19, ResNet50 deep learning models. During this period, the network’s performance was tried to be measured, and the accuracy and loss values in these calculations were shown graphically.

The human face conveys a significant amount of information. Through facial expressions, the face is able to communicate numerous sentiments without the need for verbalisation. Visual emotion recognition has been extensively studied. Recently several end-to-end trained deep neural networks have been proposed for this task. However, such models often lack generalisation ability across datasets. In this paper, we propose the Deep Facial Expression Vector ExtractoR (DeepFEVER), a new deep learning-based approach that learns a visual feature extractor general enough to be applied to any other facial emotion recognition task or dataset. DeepFEVER outperforms state-of-the-art results on the AffectNet and Google Facial Expression Comparison datasets. DeepFEVER’s extracted features also generalise extremely well to other datasets – even those unseen during training – namely, the Real-World Affective Faces (RAF) dataset.

With the development of artificial intelligence, facial expression recognition (FER) has greatly improved performance in deep learning, but there is still a lot of room for improvement in the study of combining attention to focus the network on key parts of the face. For facial expression recognition, this paper designs a network model, which use spatial transformer network to transform the input image firstly, and then adding channel attention and spatial attention to the convolutional network. In addition, in this paper, the GELU activation function is used in the convolutional network, which improves the recognition rate of facial expressions to a certain extent.

Deep learning models have been successful and shown to perform better in terms of accuracy and efficiency for facial recognition applications. However, they require huge amount of data samples that were well annotated to be successful. Their data requirements have led to some complications which include increased processing demands of the systems where such systems were to be deployed. Reducing the training sample sizes of deep learning models is still an open problem. This paper proposes the reduction of the number of samples required by the convolutional neutral network used in training a facial recognition system using a new Facial Image Deviation Estimation and Image Selection Algorithm (FIDE-ISA). The algorithm was used to select appropriate facial image training samples incrementally based on their facial deviation. This will reduce the need for huge dataset in training deep learning models. Preliminary results indicated a 100% accuracy for models trained with 54 images (at least 3 images per individual) and above.

State estimation algorithm ensures an effective realtime monitoring of the modern smart grid leading to an accurate determination of the current operating states. Recently, a new genre of data integrity attacks namely false data injection attack (FDIA) has shown its deleterious effects by bypassing the traditional bad data detection technique. Modern grid operators must detect the presence of such attacks in the raw field measurements to guarantee a safe and reliable operation of the grid. State forecasting based FDIA identification schemes have recently shown its efficacy by determining the deviation of the estimated states due to an attack. This work emphasizes on a scalable deep learning state forecasting model which can accurately determine the presence of FDIA in real-time. An optimal set of hyper-parameters of the proposed architecture leads to an effective forecasting of the operating states with minimal error. A diligent comparison between other state of the art forecasting strategies have promoted the effectiveness of the proposed neural network. A comprehensive analysis on the IEEE 14 bus test bench effectively promotes the proposed real-time attack identification strategy.

In recent years, cyber attackers are targeting the power system and imposing different damages to the national economy and public safety. False Data Injection Attack (FDIA) is one of the main types of Cyber-Physical attacks that adversaries can manipulate power system measurements and modify system data. Consequently, it may result in incorrect decision-making and control operations and lead to devastating effects. In this paper, we propose a two-stage detection method. In the first step, Gated Recurrent Unit (GRU), as a deep learning algorithm, is employed to forecast the data for the future horizon. Meanwhile, hyperparameter optimization is implemented to find the optimum parameters (i.e., number of layers, epoch, batch size, β1, β2, etc.) in the supervised learning process. In the second step, an unsupervised scoring algorithm is employed to find the sequences of false data. Furthermore, two penalty factors are defined to prevent the objective function from greedy behavior. We assess the capability of the proposed false data detection method through simulation studies on a real-world data set (ComEd. dataset, Northern Illinois, USA). The results demonstrate that the proposed method can detect different types of attacks, i.e., scaling, simple ramp, professional ramp, and random attacks, with good performance metrics (i.e., recall, precision, F1 Score). Furthermore, the proposed deep learning method can mitigate false data with the estimated true values.

Most of the studies of the existing object detection models are studies to better detect the objects to be detected. The problem of false detection of objects that should not be detected is not considered. When an object detection model that does not take this problem into account is applied to an industrial field close to humans, false detection can lead to a dangerous situation that greatly interferes with human life. To solve this false detection problem, this paper proposes a method of fine-tuning the backbone neural network model of the object detection model using the Outlier Exposure method and applying the class-specific uncertainty constant to the confidence score to detect the object.

Machine learning offers performance improvements and novel functionality, but its life cycle performance is understudied. In areas like cognitive communications, where systems are long-lived, life cycle trade-offs are key to system design. Herein, we consider the use of deep learning to classify spectrograms. We vary the label-space over which the network makes classifications, as may emerge with changes in use over a system’s life cycle, and compare heterogeneous transfer learning performance across label-spaces between model architectures. Our results offer an empirical example of life cycle challenges to using machine learning for cognitive communications. They evidence important trade-offs among performance, training time, and sensitivity to the order in which the label-space is changed. And they show that fine-tuning can be used in the heterogeneous transfer of spectrogram classifiers.

In recent years web applications that are hacked every day estimated to be 30 000, and in most cases, web developers or website owners do not even have enough knowledge about what is happening on their sites. Web hackers can use many attacks to gain entry or compromise legitimate web applications, they can also deceive people by using phishing sites to collect their sensitive and private information. In response to this, the need is raised to take proper measures to understand the risks and be aware of the vulnerabilities that may affect the website and hence the normal business flow. In the scope of this study, mitigations against the most common web application attacks are set, and the web administrator is provided with ways to detect phishing links which is a social engineering attack, the study also demonstrates the generation of web application logs that simplifies the process of analyzing the actions of abnormal users to show when behavior is out of bounds, out of scope, or against the rules. The methods of mitigation are accomplished by secure coding techniques and the methods for phishing link detection are performed by various machine learning algorithms and deep learning techniques. The developed application has been tested and evaluated against various attack scenarios, the outcomes obtained from the test process showed that the website had successfully mitigated these dangerous web application attacks, and for the detection of phishing links part, a comparison is made between different algorithms to find the best one, and the outcome of the best model gave 98% accuracy.

Deep learning (DL) is becoming popular as a new tool for many applications in wireless communication systems. However, for many classification tasks (e.g., modulation classification) it has been shown that DL-based wireless systems are susceptible to adversarial examples; adversarial examples are well-crafted malicious inputs to the neural network (NN) with the objective to cause erroneous outputs. In this paper, we extend this to regression problems and show that adversarial attacks can break DL-based power allocation in the downlink of a massive multiple-input-multiple-output (maMIMO) network. Specifically, we extend the fast gradient sign method (FGSM), momentum iterative FGSM, and projected gradient descent adversarial attacks in the context of power allocation in a maMIMO system. We benchmark the performance of these attacks and show that with a small perturbation in the input of the NN, the white-box attacks can result in infeasible solutions up to 86%. Furthermore, we investigate the performance of black-box attacks. All the evaluations conducted in this work are based on an open dataset and NN models, which are publicly available.

Multimedia Internet of Things (IoT) devices, especially, the smartphones are embedded with sensors including Global Positioning System (GPS), barometer, microphone, accelerometer, etc. These sensors working together, present a fairly complete picture of the citizens' daily activities, with implications for their privacy. With the internet, Citizens in Smart Cities are able to perform their daily life activities online with their connected electronic devices. But, unfortunately, computer hackers tend to write automated malicious applications to attack websites on which these citizens perform their activities. These security threats sometime put their private information at risk. In order to prevent these security threats on websites, Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHAs) are generated, as a form of security mechanism to protect the citizens' private information. But with the advancement of deep learning, text-based CAPTCHAs can sometimes be vulnerable. As a result, it is essential to conduct performance evaluation on the CAPTCHAs that are generated before they are deployed on multimedia web applications. Therefore, this work proposed an optimal and light-weight Convolutional Neural Network (CNN) to solve both numerical and alpha-numerical complex text-based CAPTCHAs simultaneously. The accuracy of the proposed CNN model has been accelerated based on Cyclical Learning Rates (CLRs) policy. The proposed CLR-CNN model achieved a high accuracy to solve both numerical and alpha-numerical text-based CAPTCHAs of 99.87% and 99.66%, respectively. In real-time, we observed that the speed of the model has increased, the model is lightweight, stable, and flexible as compared to other CAPTCHA solving techniques. The result of this current work will increase awareness and will assist multimedia security Researchers to continue and develop more robust text-based CAPTCHAs with their security mechanisms capable of protecting the private information of citizens in Smart Cities.