Biblio

Aiming at group shilling attacks in recommender systems, a shilling group detection approach based on triangle dense subgraph mining is proposed. First, the user relation graph is built by mining the relations among users in the rating dataset. Second, the improved triangle dense subgraph mining method and the personalizing PageRank seed expansion algorithm are used to divide candidate shilling groups. Finally, the suspicious degrees of candidate groups are calculated using several group detection indicators and the attack groups are obtained. Experiments indicate that our method has better detection performance on the Amazon and Yelp datasets than the baselines.

Symbiotic radio (SR) has recently emerged as a promising technology to boost spectrum efficiency of wireless communications by allowing reflective communications underlying the active RF communications. In this paper, we leverage SR to boost physical layer security by using an array of passive reflecting elements constituting the intelligent reflecting surface (IRS), which is reconfigurable to induce diverse RF radiation patterns. In particular, by switching the IRS's phase shifting matrices, we can proactively create dynamic channel conditions, which can be exploited by the transceivers to extract common channel features and thus used to generate secret keys for encrypted data transmissions. As such, we firstly present the design principles for IRS-assisted key generation and verify a performance improvement in terms of the secret key generation rate (KGR). Our analysis reveals that the IRS's random phase shifting may result in a non-uniform channel distribution that limits the KGR. Therefore, to maximize the KGR, we propose both a heuristic scheme and deep reinforcement learning (DRL) to control the switching of the IRS's phase shifting matrices. Simulation results show that the DRL approach for IRS-assisted key generation can significantly improve the KGR.

In the era of rapid technological growth, malicious traffic has drawn increased attention. Most well-known offensive security assessment todays are heavily focused on pre-compromise. The amount of anomalous data in today's context is massive. Analyzing the data using primitive methods would be highly challenging. Solution to it is: If we can detect adversary behaviors in the early stage of compromise, one can prevent and safeguard themselves from various attacks including ransomwares and Zero-day attacks. Integration of new technologies Artificial Intelligence & Machine Learning with manual Anomaly Detection can provide automated machine-based detection which in return can provide the fast, error free, simplify & scalable Threat Detection & Response System. Endpoint Detection & Response (EDR) tools provide a unified view of complex intrusions using known adversarial behaviors to identify intrusion events. We have used the EMBER dataset, which is a labelled benchmark dataset. It is used to train machine learning models to detect malicious portable executable files. This dataset consists of features derived from 1.1 million binary files: 900,000 training samples among which 300,000 were malicious, 300,000 were benevolent, 300,000 un-labelled, and 200,000 evaluation samples among which 100K were malicious, 100K were benign. We have also included open-source code for extracting features from additional binaries, enabling the addition of additional sample features to the dataset.

Threats, posed by ransomware, are rapidly increasing, and its cost on both national and global scales is becoming significantly high as evidenced by the recent events. Ransomware carries out an irreversible process, where it encrypts victims' digital assets to seek financial compensations. Adversaries utilize different means to gain initial access to the target machines, such as phishing emails, vulnerable public-facing software, Remote Desktop Protocol (RDP), brute-force attacks, and stolen accounts. To combat these threats of ransomware, this paper aims to help researchers gain a better understanding of ransomware application profiles through static analysis, where we identify a list of suspicious indicators and similarities among 727 active ran-somware samples. We start with generating portable executable (PE) metadata for all the studied samples. With our domain knowledge and exploratory data analysis tasks, we introduce some of the suspicious indicators of the structure of ransomware files. We reduce the dimensionality of the generated dataset by using the Principal Component Analysis (PCA) technique and discover clusters by applying the KMeans algorithm. This motivates us to utilize the one-class classification algorithms on the generated dataset. As a result, the algorithms learn the common data boundary in the structure of our studied ransomware samples, and thereby, we achieve the data-driven similarities. We use the findings to evaluate the trained classifiers with the test samples and observe that the Local Outlier Factor (LoF) performs better on all the selected feature spaces compared to the One-Class SVM and the Isolation Forest algorithms.

The number of prominent ransomware attacks has increased recently. In this research, we detect ransomware by analyzing network traffic by using machine learning algorithms and comparing their detection performances. We have developed multi-class classification models to detect families of ransomware by using the selected network traffic features, which focus on the Transmission Control Protocol (TCP). Our experiment showed that decision trees performed best for classifying ransomware families with 99.83% accuracy, which is slightly better than the random forest algorithm with 99.61% accuracy. The experimental result without feature selection classified six ransomware families with high accuracy. On the other hand, classifiers with feature selection gave nearly the same result as those without feature selection. However, using feature selection gives the advantage of lower memory usage and reduced processing time, thereby increasing speed. We discovered the following ten important features for detecting ransomware: time delta, frame length, IP length, IP destination, IP source, TCP length, TCP sequence, TCP next sequence, TCP header length, and TCP initial round trip.

Ransomware is a major malware attack experienced by large corporations and healthcare services. Ransomware employs the idea of cryptovirology, which uses cryptography to design malware. The goal of ransomware is to extort ransom by threatening the victim with the destruction of their data. Ransomware typically involves a 3-step process: analyzing the victim’s network traffic, identifying a vulnerability, and then exploiting it. Thus, the detection of ransomware has become an important undertaking that involves various sophisticated solutions for improving security. To further enhance ransomware detection capabilities, this paper focuses on an Application Programming Interface (API)-based ransomware detection approach in combination with machine learning (ML) techniques. The focus of this research is (i) understanding the life cycle of ransomware on the Windows platform, (ii) dynamic analysis of ransomware samples to extract various features of malicious code patterns, and (iii) developing and validating machine learning-based ransomware detection models on different ransomware and benign samples. Data were collected from publicly available repositories and subjected to sandbox analysis for sampling. The sampled datasets were applied to build machine learning models. The grid search hyperparameter optimization algorithm was employed to obtain the best fit model; the results were cross-validated with the testing datasets. This analysis yielded a high ransomware detection accuracy of 99.18% for Windows-based platforms and shows the potential for achieving high-accuracy ransomware detection capabilities when using a combination of API calls and an ML model. This approach can be further utilized with existing multilayer security solutions to protect critical data from ransomware attacks.

Cyberattacks have been progressed in the fields of Internet of Things, and artificial intelligence technologies using the advanced persistent threat (APT) method recently. The damage caused by ransomware is rapidly spreading among APT attacks, and the range of the damages of individuals, corporations, public institutions, and even governments are increasing. The seriousness of the problem has increased because ransomware has been evolving into an intelligent ransomware attack that spreads over the network to infect multiple users simultaneously. This study used open source endpoint detection and response tools to build and test a framework environment that enables systematic ransomware detection at the network and system level. Experimental results demonstrate that the use of EDR tools can quickly extract ransomware attack features and respond to attacks.

At present, in the face of the huge and complex data in cloud computing, the parallel computing ability of quantum computing is particularly important. Quantum principal component analysis algorithm is used as a method of quantum state tomography. We perform feature extraction on the eigenvalue matrix of the density matrix after feature decomposition to achieve dimensionality reduction, proposed quantum principal component extraction algorithm (QPCE). Compared with the classic algorithm, this algorithm achieves an exponential speedup under certain conditions. The specific realization of the quantum circuit is given. And considering the limited computing power of the client, we propose a quantum homomorphic ciphertext dimension reduction scheme (QHEDR), the client can encrypt the quantum data and upload it to the cloud for computing. And through the quantum homomorphic encryption scheme to ensure security. After the calculation is completed, the client updates the key locally and decrypts the ciphertext result. We have implemented a quantum ciphertext dimensionality reduction scheme implemented in the quantum cloud, which does not require interaction and ensures safety. In addition, we have carried out experimental verification on the QPCE algorithm on IBM's real computing platform. Experimental results show that the algorithm can perform ciphertext dimension reduction safely and effectively.

Because of the rise of the Monroe coin, many JavaScript files with embedded malicious code are used to mine cryptocurrency using the computing power of the browser client. This kind of script does not have any obvious behaviors when it is running, so it is difficult for common users to witness them easily. This feature could lead the browser side cryptocurrency mining abused without the user’s permission. Traditional browser security strategies focus on information disclosure and malicious code execution, but not suitable for such scenes. Thus, we present a novel detection method named MineDetector using a machine learning algorithm and static features for automatically detecting browser-side cryptojacking scripts on the websites. MineDetector extracts five static feature groups available from the abstract syntax tree and text of codes and combines them using the machine learning method to build a powerful cryptojacking classifier. In the real experiment, MineDetector achieves the accuracy of 99.41% and the recall of 93.55% and has better performance in time comparing with present dynamic methods. We also made our work user-friendly by developing a browser extension that is click-to-run on the Chrome browser.

Bitcoin and other digital cryptocurrencies have de-veloped rapidly in recent years. To reduce hardware and power costs, many criminals use the botnet to infect other hosts to mine cryptocurrency for themselves, which has led to the proliferation of mining botnets and is referred to as cryptojacking. At present, the mechanisms specific to cryptojacking detection include host-based, Deep Packet Inspection (DPI) based, and dynamic network characteristics based. Host-based detection requires detection installation and running at each host, and the other two are heavyweight. Besides, DPI-based detection is a breach of privacy and loses efficacy if encountering encrypted traffic. This paper de-signs a lightweight cryptojacking traffic detection method based on network behavior features for an ISP, without referring to the payload of network traffic. We set up an environment to collect cryptojacking traffic and conduct a cryptojacking traffic study to obtain its discriminative network traffic features extracted from only the first four packets in a flow. Our experimental study suggests that the machine learning classifier, random forest, based on the extracted discriminative network traffic features can accurately and efficiently detect cryptojacking traffic.

Facial expression recognition has long been established as a subject of continuous research in various fields. In this study, feature extraction was conducted by calculating the distance between facial landmarks in an image. The extracted features of the relationship between each landmark and analysis were used to classify five facial expressions. We increased the data and label reliability based on our labeling work with multiple observers. Additionally, faces were recognized from the original data, and landmark coordinates were extracted and used as features. A genetic algorithm was used to select features that were relatively more helpful for classification. We performed facial recognition classification and analysis using the method proposed in this study, which showed the validity and effectiveness of the proposed method.

Aiming at the problems of the traditional convolutional neural network (CNN), such as too many parameters, single scale feature and inefficiency by some useless features, a lightweight multi-scale network with attention is proposed for facial expression recognition. The network uses the lightweight convolutional neural network model Xception and combines with the convolutional block attention module (CBAM) to learn key facial features; In addition, depthwise separable convolution module with convolution kernel of 3 × 3, 5 × 5 and 7 × 7 are used to extract features of facial expression image, and the features are fused to expand the receptive field and obtain more rich facial feature information. Experiments on facial expression datasets Fer2013 and KDEF show that the expression recognition accuracy is improved by 2.14% and 2.18% than the original Xception model, and the results further verify the effectiveness of our methods.

Emotion recognition in the field of human-computer interaction refers to that the computer has the corresponding perceptual ability to predict the emotional state of human beings in advance by observing human expressions, behaviors and emotions, so as to ensure that computers can communicate emotionally with humans. The main research work of this paper is to extract facial image features by using Linear Discriminant Analysis (LDA) and Facial Landmark Detection after grayscale processing and cropping, and then compare the accuracy after emotion recognition and classification to determine which feature extraction method is more effective. The test results show that the accuracy rate of emotion recognition in face images can reach 73.9% by using LDA method, and 84.5% by using Facial Landmark Detection method. Therefore, facial landmarks can be used to identify emotion in face images more accurately.

Teaching evaluation is an indispensable key link in the modern education model. Its purpose is to promote learners' cognitive and non-cognitive development, especially emotional development. However, today's education has increasingly neglected the emotional process of learners' learning. Therefore, a method of using machines to analyze the emotional changes of learners during learning has been proposed. At present, most of the existing emotion recognition algorithms use the extraction of two-dimensional facial features from images to perform emotion prediction. Through research, it is found that the recognition rate of 2D facial feature extraction is not optimal, so this paper proposes an effective the algorithm obtains a single two-dimensional image from the input end and constructs a three-dimensional face model from the output end, thereby using 3D facial information to estimate the continuous emotion of the dimensional space and applying this method to an online learning system. Experimental results show that the algorithm has strong robustness and recognition ability.

As a typical cyber-physical system, the power system utilizes advanced information and communication technologies to transmit crucial control signals in communication channels. However, many adversaries can construct false data injection attacks (FDIA) to circumvent traditional bad data detection and break the stability of the power grid. In this paper, we proposed a threat-maximizing FDIA model from the view of attackers. The proposed FDIA can not only circumvent bad data detection but can also cause a terrible fluctuation in the power system. Furthermore, in order to eliminate potential attack threats, the Spatio-temporal correlations of measurement matrices are considered. To extract the Spatio-temporal features, a data-driven detection method using a deep convolutional neural network was proposed. The effectiveness of the proposed FDIA model and detection are assessed by a simulation on the New England 39 bus system. The results show that the FDIA can cause a negative effect on the power system’s stable operation. Besides, the results reveal that the proposed FDIA detection method has an outstanding performance on Spatio-temporal features extraction and FDIA recognition.

Reversible data hiding (RDH) is an emerging area in the field of information security. The RDH schemes are widely explored in the field of cloud computing for data authentication and in medical image transmission for clinical data transmission along with medical images. The RDH schemes allow the data hider to embed sensitive information in digital content in such a way that later it can be extracted while recovering the original image. In this research, we explored the use of the RDH through the encryption scheme in a biometric authentication system. The internet of things (IoT) enabled biometric authentication systems are very common nowadays. In general, in biometric authentication, computationally complex tasks such as feature extraction and feature matching will be performed in a cloud server. The user-side devices will capture biometric data such as the face, fingerprint, or iris and it will be directly communicated to the cloud server for further processing. Since the confidentiality of biometric data needs to be maintained during the transmission, the original biometric data will be encrypted using any one of the data encryption techniques. In this manuscript, we propose the use of RDH through encryption approach to transmit two different biometric data as a single file without compromising confidentiality. The proposed scheme will ensure the integrity of the biometric data during transmission. For data hiding purposes, we have used a block-wise RDH through encryption scheme. The experimental study of the proposed scheme is carried out by embedding fingerprint data in the face images. The validation of the proposed scheme is carried out by extracting the fingerprint details from the face images during image decryption. The scheme ensures the exact recovery of face image images and fingerprint data at the receiver site.

Adversarial machine learning, a technique which seeks to deceive machine learning (ML) models, threatens the utility and reliability of ML systems. This is particularly relevant in critical ML implementations such as those found in Network Intrusion Detection Systems (NIDS). This paper considers the impact of adversarial influence on NIDS and proposes ways to improve ML based systems. Specifically, we consider five feature robustness metrics to determine which features in a model are most vulnerable, and four defense methods. These methods are tested on six ML models with four adversarial sample generation techniques. Our results show that across different models and adversarial generation techniques, there is limited consistency in vulnerable features or in effectiveness of defense method.

We present a simple approach for detecting brute force attacks in the CSE-CIC-IDS2018 Big Data dataset. We show our approach is preferable to more complex approaches since it is simpler, and yields stronger classification performance. Our contribution is to show that it is possible to train and test simple Decision Tree models with two independent variables to classify CSE-CIC-IDS2018 data with better results than reported in previous research, where more complex Deep Learning models are employed. Moreover, we show that Decision Tree models trained on data with two independent variables perform similarly to Decision Tree models trained on a larger number independent variables. Our experiments reveal that simple models, with AUC and AUPRC scores greater than 0.99, are capable of detecting brute force attacks in CSE-CIC-IDS2018. To the best of our knowledge, these are the strongest performance metrics published for the machine learning task of detecting these types of attacks. Furthermore, the simplicity of our approach, combined with its strong performance, makes it an appealing technique.

Throughout the years, honeypots have been very useful in tracking down attackers and preventing different types of cyber attacks on a very large scale. It's been almost 3 decades since the discover of honeypots and still more than 80% of the companies rely on this system because of intrusion detection features and low false positive rate. But with time, the attackers tend to start discovering loopholes in the system. Hence it is very important to be up to date with the technology when it comes to protecting a computing device from the emerging cyber attacks. Timely advancements in the security model provided by the honeypots helps in a more efficient use of the resource and also leads to better innovations in that field. The following paper reviews different methods of honeypot network and also gives an insight about the problems that those techniques can face along with their solution. Further it also gives the detail about the most preferred solution among all of the listed techniques in the paper.

This paper proposes an efficient face template protection system based on homomorphic encryption. By developing a message packing method suitable for the calculation of the squared Euclidean distance, the proposed system computes the squared Euclidean distance between facial features by a single homomorphic multiplication. Our experimental results show the transaction time of the proposed system is about 14 times faster than that of the existing face template protection system based on homomorphic encryption presented in BIOSIG2020.

In recent years, hashing algorithm has been widely researched and has made considerable progress in large-scale image retrieval tasks due to its advantages of convenient storage and fast calculation efficiency. Nowadays most researchers use deep convolutional neural networks (CNNs) to perform feature learning and hash coding learning at the same time for image retrieval and the deep hashing methods based on deep CNNs perform much better than the traditional manual feature hashing methods. But most methods are designed to handle simple binary similarity and decrease quantization error, ignoring that the features of similar images and hashing codes generated are not compact enough. In order to enhance the performance of CNNs-based hashing algorithms for large scale image retrieval, this paper proposes a new deep-supervised hashing algorithm in which a novel channel attention mechanism is added and the loss function is elaborately redesigned to generate compact binary codes. It experimentally proves that, compared with the existing hashing methods, this method has better performance on two large scale image datasets CIFAR-10 and NUS-WIDE.

The Border Gateway Protocol (BGP) is in charge of the route exchange at the Internet scale. Anomalies in BGP can have several causes (mis-configuration, outage and attacks). These anomalies are classified into large or small scale anomalies. Machine learning models are used to analyze and detect anomalies from the complex data extracted from BGP behavior. Two types of data representation can be used inside the machine learning models: a graph representation of the network (graph features) or a statistical computation on the data (statistical features). In this paper, we evaluate and compare the accuracy of machine learning models using graph features and statistical features on both large and small scale BGP anomalies. We show that statistical features have better accuracy for large scale anomalies, and graph features increase the detection accuracy by 15% for small scale anomalies and are well suited for BGP small scale anomaly detection.

In order to improve the security and anti-interference ability of industrial control system, this paper proposes an improved industrial neural network defense method based on the PCA dimension reduction and the improved deep neural network. Firstly, the proposed method reduces the dimensionality of the industrial data using the dimension reduction theory of principal component analysis (PCA). Then the deep neural network extracts the features of the network. Finally, the softmax classifier classifies industrial data. Experiment results show that compared with unintegrated algorithm, this method achieves higher recognition accuracy and has great application potential.

Algorithms for unsupervised anomaly detection have proven their effectiveness and flexibility, however, first it is necessary to calculate with what ratio a certain class begins to be considered anomalous by the autoencoder. For this reason, we propose to conduct a study of the efficiency of autoencoders depending on the ratio of anomalous and non-anomalous classes. The emergence of high-speed networks in electric power systems creates a tight interaction of cyberinfrastructure with the physical infrastructure and makes the power system susceptible to cyber penetration and attacks. To address this problem, this paper proposes an innovative approach to develop a specification-based intrusion detection framework that leverages available information provided by components in a contemporary power system. An autoencoder is used to encode the causal relations among the available information to create patterns with temporal state transitions, which are used as features in the proposed intrusion detection. This allows the proposed method to detect anomalies and cyber attacks.

A new round of scientific and technological revolution and industrial reform promote the intelligent development of automobile and promote the deep integration of automobile with Internet, big data, communication and other industries. At the same time, it also brings network and data security problems to automobile, which is very easy to cause national security and social security risks. Intelligent vehicle Ethernet intrusion detection can effectively alleviate the security risk of vehicle network, but the complex attack means and vehicle compatibility have not been effectively solved. This research takes the vehicle Ethernet as the research object, constructs the machine learning samples for neural network, applies the self coding network technology combined with the original characteristics to the network intrusion detection algorithm, and studies a self-learning vehicle Ethernet intrusion detection algorithm. Through the application and test of vehicle terminal, the algorithm generated in this study can be used for vehicle terminal with Ethernet communication function, and can effectively resist 34 kinds of network attacks in four categories. This method effectively improves the network security defense capability of vehicle Ethernet, provides technical support for the network security of intelligent vehicles, and can be widely used in mass-produced intelligent vehicles with Ethernet.