Biblio

False news has become widespread in the last decade in political, economic, and social dimensions. This has been aided by the deep entrenchment of social media networking in these dimensions. Facebook and Twitter have been known to influence the behavior of people significantly. People rely on news/information posted on their favorite social media sites to make purchase decisions. Also, news posted on mainstream and social media platforms has a significant impact on a particular country’s economic stability and social tranquility. Therefore, there is a need to develop a deceptive system that evaluates the news to avoid the repercussions resulting from the rapid dispersion of fake news on social media platforms and other online platforms. To achieve this, the proposed system uses the preprocessing stage results to assign specific vectors to words. Each vector assigned to a word represents an intrinsic characteristic of the word. The resulting word vectors are then applied to RNN models before proceeding to the LSTM model. The output of the LSTM is used to determine whether the news article/piece is fake or otherwise.

In this decade, digital transactions have risen exponentially demanding more reliable and secure authentication systems. CAPTCHA (Completely Automated Public Turing Test to tell Computers and Humans Apart) system plays a major role in these systems. These CAPTCHAs are available in character sequence, picture-based, and audio-based formats. It is very essential that these CAPTCHAs should be able to differentiate a computer program from a human precisely. This work tests the strength of text-based CAPTCHAs by breaking them using an algorithm built on CNN (Convolution Neural Network) and RNN (Recurrent Neural Network). The algorithm is designed in such a way as an attempt to break the security features designers have included in the CAPTCHAs to make them hard to be cracked by machines. This algorithm is tested against the synthetic dataset generated in accordance with the schemes used in popular websites. The experiment results exhibit that the model has shown a considerable performance against both the synthetic and real-world CAPTCHAs.

This paper presents CaptchaGG, a model for recognizing linear graphical CAPTCHAs. As in the previous society, CAPTCHA is becoming more and more complex, but in some scenarios, complex CAPTCHA is not needed, and usually, linear graphical CAPTCHA can meet the corresponding functional scenarios, such as message boards of websites and registration of accounts with low security. The scheme is based on convolutional neural networks for feature extraction of CAPTCHAs, recurrent neural forests A neural network that is too complex will lead to problems such as difficulty in training and gradient disappearance, and too simple will lead to underfitting of the model. For the single problem of linear graphical CAPTCHA recognition, the model which has a simple architecture, extracting features by convolutional neural network, sequence modeling by recurrent neural network, and finally classification and recognition, can achieve an accuracy of 96% or more recognition at a lower complexity.

Security is undoubtedly the most serious problem for Web applications, and SQL injection (SQLi) attacks are one of the most damaging. The detection of SQL blind injection vulnerability is very important, but unfortunately, it is not fast enough. This is because time-based SQL blind injection lacks web page feedback, so the delay function can only be set artificially to judge whether the injection is successful by observing the response time of the page. However, brute force cracking and binary search methods used in injection require more web requests, resulting in a long time to obtain database information in SQL blind injection. In this paper, a gated recurrent neural network-based SQL blind injection technology is proposed to generate the predictive characters in SQL blind injection. By using the neural language model based on deep learning and character sequence prediction, the method proposed in this paper can learn the regularity of common database information, so that it can predict the next possible character according to the currently obtained database information, and sort it according to probability. In this paper, the training model is evaluated, and experiments are carried out on the shooting range to compare the method used in this paper with sqlmap (the most advanced sqli test automation tool at present). The experimental results show that the method used in this paper is more effective and significant than sqlmap in time-based SQL blind injection. It can obtain the database information of the target site through fewer requests, and run faster.

To improve efficiency and resource usage in data retrieval, an Internet of Things (IoT) search engine organizes a vast amount of scattered data and responds to client queries with processed results. Machine learning provides a deep understanding of complex patterns and enables enhanced feedback to users through well-trained models. Nonetheless, machine learning models are prone to adversarial attacks via the injection of elaborate perturbations, resulting in subverted outputs. Particularly, adversarial attacks on time-series data demand urgent attention, as sensors in IoT systems are collecting an increasing volume of sequential data. This paper investigates adversarial attacks on time-series analysis in an IoT search engine (IoTSE) system. Specifically, we consider the Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) as our base model, implemented in a simulated federated learning scheme. We propose the Federated Adversarial Learning for IoT Search Engine (FALIoTSE) that exploits the shared parameters of the federated model as the target for adversarial example generation and resiliency. Using a real-world smart parking garage dataset, the impact of an attack on FALIoTSE is demonstrated under various levels of perturbation. The experiments show that the training error increases significantly with noises from the gradient.

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.

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.

This paper presents conceptual modelling of the criticality of critical infrastructure (CI) nth order dependency effect using neural networks. Incidentally, critical infrastructures are usually not stand-alone, they are mostly interconnected in some way thereby creating a complex network of infrastructures that depend on each other. The relationships between these infrastructures can be either unidirectional or bidirectional with possible cascading or escalating effect. Moreover, the dependency relationships can take an nth order, meaning that a failure or disruption in one infrastructure can cascade to nth interconnected infrastructure. The nth-order dependency and criticality problems depict a sequential characteristic, which can result in chronological cyber effects. Consequently, quantifying the criticality of infrastructure demands that the impact of its failure or disruption on other interconnected infrastructures be measured effectively. To understand the complex relational behaviour of nth order relationships between infrastructures, we model the behaviour of nth order dependency using Neural Network (NN) to analyse the degree of dependency and criticality of the dependent infrastructure. The outcome, which is to quantify the Criticality Index Factor (CIF) of a particular infrastructure as a measure of its risk factor can facilitate a collective response in the event of failure or disruption. Using our novel NN approach, a comparative view of CIFs of infrastructures or organisations can provide an efficient mechanism for Critical Information Infrastructure Protection and resilience (CIIPR) in a more coordinated and harmonised way nationally. Our model demonstrates the capability to measure and establish the degree of dependency (or interdependency) and criticality of CIs as a criterion for a proactive CIIPR.

Although several different attacks or modern security mechanisms have been proposed, the captchas created by the numbers and the letters are still used by some websites or applications to protect their information security. The reason is that the labels of the captcha data are difficult to collect for the attacker, and protector can easily control the various parameters of the captchas: like the noise, the font type, the font size, and the background color, then make this security mechanism update with the increased attack methods. It can against attacks in different situations very effectively. This paper presents a method to recognize the different text-based captchas based on a system constituted by the denoising autoencoder and the Convolutional Recurrent Neural Network (CRNN) model with the Connectionist Temporal Classification (CTC) structure. We show that our approach has a better performance for recognizing, and it solves the identification problem of indefinite character length captchas efficiently.

Bitcoin has received a lot of attention from investors, researchers, regulators, and the media. It is a known fact that the Bitcoin price usually fluctuates greatly. However, not enough scientific research has been done on these fluctuations. In this study, long short-term memory (LSTM) modeling from Recurrent Neural Networks, which is one of the deep learning methods, was applied on Bitcoin values. As a result of this application, anomaly detection was carried out in the values from the data set. With the LSTM network, a time-dependent representation of Bitcoin price can be captured, and anomalies can be selected. The factors that play a role in the formation of the model to be applied in the detection of anomalies with the experimental results were evaluated.

Insider threat makes enterprises or organizations suffer from the loss of property and the negative influence of reputation. User behavior analysis is the mainstream method of insider threat detection, but due to the lack of fine-grained detection and the inability to effectively capture the behavior patterns of individual users, the accuracy and precision of detection are insufficient. To solve this problem, this paper designs an insider threat detection method based on user historical behavior and attention mechanism, including using Long Short Term Memory (LSTM) to extract user behavior sequence information, using Attention-based on user history behavior (ABUHB) learns the differences between different user behaviors, uses Bidirectional-LSTM (Bi-LSTM) to learn the evolution of different user behavior patterns, and finally realizes fine-grained user abnormal behavior detection. To evaluate the effectiveness of this method, experiments are conducted on the CMU-CERT Insider Threat Dataset. The experimental results show that the effectiveness of this method is 3.1% to 6.3% higher than that of other comparative model methods, and it can detect insider threats in different user behaviors with fine granularity.

Currently, more and more malicious insiders are making threats, and the detection of insider threats is becoming more challenging. The malicious insider often uses legitimate access privileges and mimic normal behaviors to evade detection, which is difficult to be detected via using traditional defensive solutions. In this paper, we propose DeepMIT, a malicious insider threat detection framework, which utilizes Recurrent Neural Network (RNN) to model user behaviors as time sequences and predict the probabilities of anomalies. This framework allows DeepMIT to continue learning, and the detections are made in real time, that is, the anomaly alerts are output as rapidly as data input. Also, our framework conducts further insight of the anomaly scores and provides the contributions to the scores and, thus, significantly helps the operators to understand anomaly scores and take further steps quickly(e.g. Block insider's activity). In addition, DeepMIT utilizes user-attributes (e.g. the personality of the user, the role of the user) as categorical features to identify the user's truly typical behavior, which help detect malicious insiders who mimic normal behaviors. Extensive experimental evaluations over a public insider threat dataset CERT (version 6.2) have demonstrated that DeepMIT has outperformed other existing malicious insider threat solutions.

We present a scalable architecture based on a trained filter bank for input pre-processing and a recurrent neural network (RNN) for the detection of atrial fibrillation in electrocardiogram (ECG) signals, with the focus on enabling a very efficient hardware implementation as application-specific integrated circuit (ASIC). Our already very efficient base architecture is further improved by replacing the RNN with a delta-encoded gated recurrent unit (GRU) and adding a confidence measure (CM) for terminating the computation as early as possible. With these optimizations, we demonstrate a reduction of the processing load of 58 % on an internal dataset while still achieving near state-of-the-art classification results on the Physionet ECG dataset with only 1202 parameters.

Sentiment analytics is the process of applying natural language processing and methods for text-based information to define and extract subjective knowledge of the text. Natural language processing and text classifications can deal with limited corpus data and more attention has been gained by semantic texts and word embedding methods. Deep learning is a powerful method that learns different layers of representations or qualities of information and produces state-of-the-art prediction results. In different applications of sentiment analytics, deep learning methods are used at the sentence, document, and aspect levels. This review paper is based on the main difficulties in the sentiment assessment stage that significantly affect sentiment score, pooling, and polarity detection. The most popular deep learning methods are a Convolution Neural Network and Recurrent Neural Network. Finally, a comparative study is made with a vast literature survey using deep learning models.

In recent years, the complexity of software vulnera-bilities has continued to increase. Manual vulnerability detection methods alone no longer meet the demand. With the rapid development of the deep learning, many neural network models have been widely applied to source code vulnerability detection. The variant of recurrent neural network (RNN), bidirectional Long Short-Term Memory (BiLSTM), has been a popular choice in vulnerability detection. However, is BiLSTM the most suitable choice? To answer this question, we conducted a series of experiments to investigate the effectiveness of different neural network models for source code vulnerability detection. The results shows that the variants of RNN, gated recurrent unit (GRU) and bidirectional GRU, are more capable of detecting source code fragments with mixed vulnerability types. And the concatenated convolutional neural network is more capable of detecting source code fragments of single vulnerability types.

Accidents in Nuclear Power Plants (NPPs) can occur for a variety of causes. However, among these, the scale of accidents due to human error can be greater than expected. Accordingly, researches are being actively conducted using artificial intelligence to reduce human error. Most of the research shows high performance based on the numerical data on NPPs, but the expandability of researches using only numerical data is limited. Therefore, in this study, abnormal diagnosis was performed using artificial intelligence based on image data. The methods applied to abnormal diagnosis are the deep neural network, convolution neural network, and convolution recurrent neural network. Consequently, in nuclear power plants, it is expected that the application of more methodologies can be expanded not only in numerical data but also in image-based data.

Edge nodes are crucial for detection against multitudes of cyber attacks on Internet-of-Things endpoints and is set to become part of a multi-billion industry. The resource constraints in this novel network infrastructure tier constricts the deployment of existing Network Intrusion Detection System with Deep Learning models (DLM). We address this issue by developing a novel light, fast and accurate `Edge-Detect' model, which detects Distributed Denial of Service attack on edge nodes using DLM techniques. Our model can work within resource restrictions i.e. low power, memory and processing capabilities, to produce accurate results at a meaningful pace. It is built by creating layers of Long Short-Term Memory or Gated Recurrent Unit based cells, which are known for their excellent representation of sequential data. We designed a practical data science pipeline with Recurring Neural Network to learn from the network packet behavior in order to identify whether it is normal or attack-oriented. The model evaluation is from deployment on actual edge node represented by Raspberry Pi using current cybersecurity dataset (UNSW2015). Our results demonstrate that in comparison to conventional DLM techniques, our model maintains a high testing accuracy of 99% even with lower resource utilization in terms of cpu and memory. In addition, it is nearly 3 times smaller in size than the state-of-art model and yet requires a much lower testing time.

Advanced persistent threats (APT’s) are stealthy threat actors with the skills to gain covert control of the computer network for an extended period of time. They are the highest cyber attack risk factor for large companies and states. A successful attack via an APT can cost millions of dollars, can disrupt civil life and has the capabilities to do physical damage. APT groups are typically state-sponsored and are considered the most effective and skilled cyber attackers. Attacks of APT’s are executed in several stages as pointed out in the Lockheed Martin cyber kill chain (CKC). Each of these APT stages can potentially be identified as patterns in network traffic. Using the "APT-2020" dataset, that compiles the characteristics and stages of an APT, we carried out experiments on the detection of anomalous traffic for all APT stages. We compare several artificial intelligence models, like a stacked auto encoder, a recurrent neural network and a one class state vector machine and show significant improvements on detection in the data exfiltration stage. This dataset is the first to have a data exfiltration stage included to experiment on. According to APT-2020’s authors current models have the biggest challenge specific to this stage. We introduce a method to successfully detect data exfiltration by analyzing the payload of the network traffic flow. This flow based deep packet inspection approach improves detection compared to other state of the art methods.

E-commerce, ticket booking, banking, and other web-based applications that deal with sensitive information, such as passwords, payment information, and financial information, are widespread. Some web developers may have different levels of understanding about securing an online application. The two vulnerabilities identified by the Open Web Application Security Project (OWASP) for its 2017 Top Ten List are SQL injection and Cross-site Scripting (XSS). Because of these two vulnerabilities, an attacker can take advantage of these flaws and launch harmful web-based actions. Many published articles concentrated on a binary classification for these attacks. This article developed a new approach for detecting SQL injection and XSS attacks using deep learning. SQL injection and XSS payloads datasets are combined into a single dataset. The word-embedding technique is utilized to convert the word’s text into a vector. Our model used BiLSTM to auto feature extraction, training, and testing the payloads dataset. BiLSTM classified the payloads into three classes: XSS, SQL injection attacks, and normal. The results showed great results in classifying payloads into three classes: XSS attacks, injection attacks, and non-malicious payloads. BiLSTM showed high performance reached 99.26% in terms of accuracy.

Automatic modulation classification (AMC) aims to improve the efficiency of crowded radio spectrums by automatically predicting the modulation constellation of wireless RF signals. Recent work has demonstrated the ability of deep learning to achieve robust AMC performance using raw in-phase and quadrature (IQ) time samples. Yet, deep learning models are highly susceptible to adversarial interference, which cause intelligent prediction models to misclassify received samples with high confidence. Furthermore, adversarial interference is often transferable, allowing an adversary to attack multiple deep learning models with a single perturbation crafted for a particular classification network. In this work, we present a novel receiver architecture consisting of deep learning models capable of withstanding transferable adversarial interference. Specifically, we show that adversarial attacks crafted to fool models trained on time-domain features are not easily transferable to models trained using frequency-domain features. In this capacity, we demonstrate classification performance improvements greater than 30% on recurrent neural networks (RNNs) and greater than 50% on convolutional neural networks (CNNs). We further demonstrate our frequency feature-based classification models to achieve accuracies greater than 99% in the absence of attacks.

The paper considers the issue of recurrent neural networks applicability for detecting industrial process anomalies to detect intrusion in Industrial Control Systems. Cyberattack on Industrial Control Systems often leads to appearing of anomalies in industrial process. Thus, it is proposed to detect such anomalies by forecasting the state of an industrial process using a recurrent neural network and comparing the predicted state with actual process' state. In the course of experimental research, a recurrent neural network with one-dimensional convolutional layer was implemented. The Secure Water Treatment dataset was used to train model and assess its quality. The obtained results indicate the possibility of using the proposed method in practice. The proposed method is characterized by the absence of the need to use anomaly data for training. Also, the method has significant interpretability and allows to localize an anomaly by pointing to a sensor or actuator whose signal does not match the model's prediction.

Application programming interfaces (APIs) are a vital part of every online business. APIs are responsible for transferring data across systems within a company or to the users through the web or mobile applications. Security is a concern for any public-facing application. The objective of this study is to analyze incoming requests to a target API and flag any malicious activity. This paper proposes a solution using sequence models to identify whether or not an API request has SQL, XML, JSON, and other types of malicious injections. We also propose a novel heuristic procedure that minimizes the number of false positives. False positives are the valid API requests that are misclassified as malicious by the model.

Insider threat is one of the most pernicious threat vectors to information and communication technologies (ICT) across the world due to the elevated level of trust and access that an insider is afforded. This type of threat can stem from both malicious users with a motive as well as negligent users who inadvertently reveal details about trade secrets, company information, or even access information to malignant players. In this paper, we propose a novel approach that uses system logs to detect insider behavior using a special recurrent neural network (RNN) model. Ground truth is established using DANTE and used as baseline for identifying anomalous behavior. For this, system logs are modeled as a natural language sequence and patterns are extracted from these sequences. We create workflows of sequences of actions that follow a natural language logic and control flow. These flows are assigned various categories of behaviors - malignant or benign. Any deviation from these sequences indicates the presence of a threat. We further classify threats into one of the five categories provided in the CERT insider threat dataset. Through experimental evaluation, we show that the proposed model can achieve 93% prediction accuracy.

Knowledge Tracing (KT) is a task that aims to assess students' mastery level of knowledge and predict their performance over questions, which has attracted widespread attention over the years. Recently, an increasing number of researches have applied deep learning techniques to knowledge tracing and have made a huge success over traditional Bayesian Knowledge Tracing methods. Most existing deep learning-based methods utilized either Recurrent Neural Networks (RNNs) or Convolutional Neural Networks (CNNs). However, it is worth noticing that these two sorts of models are complementary in modeling abilities. Thus, in this paper, we propose a novel knowledge tracing model by taking advantage of both two models via combining them into a single integrated model, named Convolutional Recurrent Knowledge Tracing (CRKT). Extensive experiments show that our model outperforms the state-of-the-art models in multiple KT datasets.

Recommendation systems, especially collaborative filtering recommenders, are vulnerable to shilling attacks as some profit-driven users may inject fake profiles into the system to alter recommendation outputs. Current shilling attack detection methods are mostly based on feature extraction techniques. The hand-designed features can confine the model to specific domains or datasets while deep learning techniques enable us to derive deeper level features, enhance detection performance, and generalize the solution on various datasets and domains. This paper illustrates the application of two deep learning methods to detect shilling attacks. We conducted experiments on the MovieLens 100K and Netflix Dataset with different levels of attacks and types. Experimental results show that deep learning models can achieve an accuracy of up to 99%.