Biblio

Adversarial attacks have recently been proposed to scrutinize the security of deep neural networks. Most blackbox adversarial attacks, which have partial access to the target through queries, are target-specific; e.g., they require a well-trained surrogate that accurately mimics a given target. In contrast, target-agnostic black-box attacks are developed to attack any target; e.g., they learn a generalized surrogate that can adapt to any target via fine-tuning on samples queried from the target. Despite their success, current state-of-the-art target-agnostic attacks require tremendous fine-tuning steps and consequently an immense number of queries to the target to generate successful attacks. The high query complexity of these attacks makes them easily detectable and thus defendable. We propose a novel query-efficient target-agnostic attack that trains a generalized surrogate network to output the adversarial directions iv.r.t. the inputs and equip it with an effective fine-tuning strategy that only fine-tunes the surrogate when it fails to provide useful directions to generate the attacks. Particularly, we show that to effectively adapt to any target and generate successful attacks, it is sufficient to fine-tune the surrogate with informative samples that help the surrogate get out of the failure mode with additional information on the target’s local behavior. Extensive experiments on CIFAR10 and CIFAR-100 datasets demonstrate that the proposed target-agnostic approach can generate highly successful attacks for any target network with very few fine-tuning steps and thus significantly smaller number of queries (reduced by several order of magnitudes) compared to the state-of-the-art baselines.

Recent studies show that the state-of-the-art deep neural networks are vulnerable to model inversion attacks, in which access to a model is abused to reconstruct private training data of any given target class. Existing attacks rely on having access to either the complete target model (whitebox) or the model's soft-labels (blackbox). However, no prior work has been done in the harder but more practical scenario, in which the attacker only has access to the model's predicted label, without a confidence measure. In this paper, we introduce an algorithm, Boundary-Repelling Model Inversion (BREP-MI), to invert private training data using only the target model's predicted labels. The key idea of our algorithm is to evaluate the model's predicted labels over a sphere and then estimate the direction to reach the target class's centroid. Using the example of face recognition, we show that the images reconstructed by BREP-MI successfully reproduce the semantics of the private training data for various datasets and target model architectures. We compare BREP-MI with the state-of-the-art white-box and blackbox model inversion attacks, and the results show that despite assuming less knowledge about the target model, BREP-MI outperforms the blackbox attack and achieves comparable results to the whitebox attack. Our code is available online.11https://github.com/m-kahla/Label-Only-Model-Inversion-Attacks-via-Boundary-Repulsion

Black-box adversarial attack has aroused much research attention for its difficulty on nearly no available information of the attacked model and the additional constraint on the query budget. A common way to improve attack efficiency is to transfer the gradient information of a white-box substitute model trained on an extra dataset. In this paper, we deal with a more practical setting where a pre-trained white-box model with network parameters is provided without extra training data. To solve the model mismatch problem between the white-box and black-box models, we propose a novel algorithm EigenBA by systematically integrating gradient-based white-box method and zeroth-order optimization in black-box methods. We theoretically show the optimal directions of perturbations for each step are closely related to the right singular vectors of the Jacobian matrix of the pretrained white-box model. Extensive experiments on ImageNet, CIFAR-10 and WebVision show that EigenBA can consistently and significantly outperform state-of-the-art baselines in terms of success rate and attack efficiency.

Classic black-box adversarial attacks can take advantage of transferable adversarial examples generated by a similar substitute model to successfully fool the target model. However, these substitute models need to be trained by target models' training data, which is hard to acquire due to privacy or transmission reasons. Recognizing the limited availability of real data for adversarial queries, recent works proposed to train substitute models in a data-free black-box scenario. However, their generative adversarial networks (GANs) based framework suffers from the convergence failure and the model collapse, resulting in low efficiency. In this paper, by rethinking the collaborative relationship between the generator and the substitute model, we design a novel black-box attack framework. The proposed method can efficiently imitate the target model through a small number of queries and achieve high attack success rate. The comprehensive experiments over six datasets demonstrate the effectiveness of our method against the state-of-the-art attacks. Especially, we conduct both label-only and probability-only attacks on the Microsoft Azure online model, and achieve a 100% attack success rate with only 0.46% query budget of the SOTA method [49].

Automatic speech recognition (ASR) models are used widely in applications for voice navigation and voice control of domestic appliances. ASRs have been misused by attackers to generate malicious outputs by attacking the deep learning component within ASRs. To assess the security and robustnesss of ASRs, we propose techniques within our framework SPAT that generate blackbox (agnostic to the DNN) adversarial attacks that are portable across ASRs. This is in contrast to existing work that focuses on whitebox attacks that are time consuming and lack portability. Our techniques generate adversarial attacks that have no human audible difference by manipulating the input speech signal using a psychoacoustic model that maintains the audio perturbations below the thresholds of human perception. We propose a framework SPAT with three attack generation techniques based on the psychoacoustic concept and frame selection techniques to selectively target the attack. We evaluate portability and effectiveness of our techniques using three popular ASRs and two input audio datasets using the metrics- Word Error Rate (WER) of output transcription, Similarity to original audio, attack Success Rate on different ASRs and Detection score by a defense system. We found our adversarial attacks were portable across ASRs, not easily detected by a state-of the-art defense system, and had significant difference in output transcriptions while sounding similar to original audio.

To detect human behaviour and measure accuracy of classification rate. Materials and Methods: A novel deep belief network with sample size 10 and support vector machine with sample size of 10. It was iterated at different times predicting the accuracy percentage of human behaviour. Results: Human behaviour detection utilizing novel deep belief network 87.9% accuracy compared with support vector machine 87.0% accuracy. Deep belief networks seem to perform essentially better compared to support vector machines \$(\textbackslashmathrmp=0.55)(\textbackslashtextPiˆ0.05)\$. The deep belief algorithm in computer vision appears to perform significantly better than the support vector machine algorithm. Conclusion: Within this human behaviour detection novel deep belief network has more precision than support vector machine.

In recent years, radar automatic target recognition (RATR) technology based on high-resolution range profile (HRRP) has received extensive attention in various fields. However, insufficient data on non-cooperative targets seriously affects recognition performance of this technique. For HRRP target recognition under few-shot condition, we proposed a novel gaussian deep belief network based on model-agnostic meta-learning (GDBN-MAML). In the proposed method, GDBN allowed real-value data to be transmitted over the entire network, which effectively avoided feature loss due to binarization requirements of conventional deep belief network (DBN) for data. In addition, we optimized the initial parameters of GDBN by multi-task learning based on MAML. In this way, the number of training samples required by the model for new recognition tasks could be reduced. We applied the proposed method to the HRRP recognition experiments of 3 types of 3D simulated aircraft models. The experimental results showed that the proposed method had higher recognition accuracy and generalization performance under few-shot condition compared with conventional deep learning methods.

Aim: Object Detection is one of the latest topics in today’s world for detection of real time objects using Deep Belief Networks. Methods & Materials: Real-Time Object Detection is performed using Deep Belief Networks (N=24) over Convolutional Neural Networks (N=24) with the split size of training and testing dataset 70% and 30% respectively. Results: Deep Belief Networks has significantly better accuracy (81.2%) compared to Convolutional Neural Networks (47.7%) and attained significance value of p = 0.083. Conclusion: Deep Belief Networks achieved significantly better object detection than Convolutional Neural Networks for identifying real-time objects in traffic surveillance.

The topological structure of the network relationship is described by the network diagram, and the formation and evolution process of the network is analyzed by using the cost-benefit method. Assuming that the self-interested network member nodes can connect or break the connection, the network topology model is established based on the dynamic random pairing evolution network model. The static structure of the network is studied. Respecting the psychological cognition law of college students and innovating the core value cultivation model can reverse the youth's identification dilemma with the core values, and then create a good political environment for the normal, healthy, civilized and orderly network participation of the youth. In recognition of the atmosphere, an automatic learning algorithm of Bayesian network structure that effectively integrates expert knowledge and data-driven methods is realized.

Historically, energy resources are of strategic importance for the social welfare and economic growth. So, predicting crude oil price fluctuations is an important issue. Since crude oil price changes are affected by many risk factors in markets, this price shows more complicated nonlinear behavior and creates more risk levels for investors than in the past. We propose a new method of prediction of crude oil price to model nonlinear dynamics. The results of the experiments show that the superior performance of the model based on the proposed method against statistical previous works is statistically significant. In general, we found that the combination of the IDBN or LSTM model lowered the MSE value to 4.65, which is 0.81 lower than the related work (Chen et al. protocol), indicating an improvement in prediction accuracy.

Higher education management has problems producing 100% of graduates capable of responding to the needs of industry while industry also is struggling to find qualified graduates that responded to their needs in part because of the inefficient way of evaluating problems, as well as because of weaknesses in the evaluation of problem-solving capabilities. The objective of this paper is to propose an appropriate classification model to be used for predicting and evaluating the attributes of the data set of the student in order to meet the selection criteria required by the industries in the academic field. The dataset required for this analysis was obtained from a private firm and the execution was carried out using Chimp Optimization Algorithm (COA) based Deep Belief Neural Network (COA-DBNN) and the obtained results are compared with various classifiers such as Logistic Regression (LR), Decision Tree (DT) and Random Forest (RF). The proposed model outperforms other classifiers in terms of various performance metrics. This critical analysis will help the college management to make a better long-term plan for producing graduates who are skilled, knowledgeable and fulfill the industry needs as well.

In view of the characteristics that rolling bearing is prone to failure under actual working conditions, and it is difficult to classify the fault category and fault degree, the deep belief network is introduced to diagnose the rolling bearing fault. Firstly, principal component analysis is used to reduce the dimension of original input data and delete redundant input information. Then, the dimension reduced data are input into the deep belief network to extract the low dimensional fault feature representation, and the extracted features are input into the classifier for rolling bearing fault pattern recognition. Finally, the diagnosis effect of the proposed network is compared with the existing common shallow neural network. The simulation experiment is carried out through the bearing data in the United States.

VCB is an important component to ensure the safe and smooth operation of the power system. As an important driving part of the vacuum circuit breaker, the operating mechanism is prone to mechanical failure, which leads to power grid accidents. This paper offers an in-depth analysis of the mechanical faults of the operating mechanism of vacuum circuit breaker and their causes, extracts the current signal of the opening and closing coil strongly correlated with the mechanical faults of the operating mechanism as the characteristic information to build a Deep Belief Network (DBN) model, trains each data set via Restricted Boltzmann Machine(RBM) and updates the model parameters. The number of hidden layer nodes, the structure of the network layer, and the learning rate are determined, and the mechanical fault diagnosis system of vacuum circuit breaker based on the Deep Belief Network is established. The results show that when the network structure is 8-110-110-6 and the learning rate is 0.01, the recognition accuracy of the DBN model is the highest, which is 0.990871. Compared with BP neural network, DBN has a smaller cross-entropy error and higher accuracy. This method can accurately diagnose the mechanical fault of the vacuum circuit breaker, which lays a foundation for the smooth operation of the power system.

The selection of distribution network faults is of great significance to accurately identify the fault location, quickly restore power and improve the reliability of power supply. This paper mainly studies the fault phase selection method of distribution network based on wavelet singular entropy and deep belief network (DBN). Firstly, the basic principles of wavelet singular entropy and DBN are analyzed, and on this basis, the DBN model of distribution network fault phase selection is proposed. Firstly, the transient fault current data of the distribution network is processed to obtain the wavelet singular entropy of the three phases, which is used as the input of the fault phase selection model; then the DBN network is improved, and an artificial neural network (ANN) is introduced to make it a fault Select the phase classifier, and specify the output label; finally, use Simulink to build a simulation model of the IEEE33 node distribution network system, obtain a large amount of data of various fault types, generate a training sample library and a test sample library, and analyze the neural network. The adjustment of the structure and the training of the parameters complete the construction of the DBN model for the fault phase selection of the distribution network.

In the operation of information technology (IT) services, operators monitor the equipment-issued alarms, to locate the cause of a failure and take action. Alarms generate simultaneously from multiple devices with physical/logical connections. Therefore, if the time and location of the alarms are close to each other, it can be judged that the alarms are likely to be caused by the same event. In this paper, we propose a method that takes a novel approach by correlating alarms considering event units using a Bayesian network based on alarm generation time, generation place, and alarm type. The topology information becomes a critical decision element when doing the alarm correlation. However, errors may occur when topology information updates manually during failures or construction. Therefore, we show that event-by-event correlation with 100% accuracy is possible even if the topology information is 25% wrong by taking into location information other than topology information.

Based on the analysis of material performance data management requirements, a network-sharing scheme of material performance data is proposed. A material performance database system including material performance data collection, data query, data analysis, data visualization, data security management and control modules is designed to solve the problems of existing material performance database network sharing, data fusion and multidisciplinary support, and intelligent services Inadequate standardization and data security control. This paper adopts hierarchical access control strategy. After logging into the material performance database system, users can standardize the material performance data and store them to form a shared material performance database. The standardized material performance data of the database system shall be queried and shared under control according to the authority. Then, the database system compares and analyzes the material performance data obtained from controlled query sharing. Finally, the database system visualizes the shared results of controlled queries and the comparative analysis results obtained. The database system adopts the MVC architecture based on B/S (client/server) cross platform J2EE. The Third-party computing platforms are integrated in System. Users can easily use material performance data and related services through browsers and networks. MongoDB database is used for data storage, supporting distributed storage and efficient query.

Data Analytics is at the core of almost all modern ap-plications ranging from science and finance to healthcare and web applications. The evolution of data analytics over the last decade has been dramatic - new methods, new tools and new platforms - with no slowdown in sight. This rapid evolution has pushed the boundaries of data analytics along several axis including scalability especially with the rise of distributed infrastructures and the Big Data era, and interoperability with diverse data management systems such as relational databases, Hadoop and Spark. However, many analytic application developers struggle with the challenge of production deployment. Recent experience suggests that it is difficult to deliver modern data analytics with the level of reliability, security and manageability that has been a feature of traditional SQL DBMSs. In this tutorial, we discuss the advances and innovations introduced at both the infrastructure and algorithmic levels, directed at making analytic workloads scale, while paying close attention to the kind of quality of service guarantees different technology provide. We start with an overview of the classical centralized analytical techniques, describing the shift towards distributed analytics over non-SQL infrastructures. We contrast such approaches with systems that integrate analytic functionality inside, above or adjacent to SQL engines. We also explore how Cloud platforms' virtualization capabilities make it easier - and cheaper - for end users to apply these new analytic techniques to their data. Finally, we conclude with the learned lessons and a vision for the near future.

Databases are at the heart of modern applications and any threats to them can seriously endanger the safety and functionality of applications relying on the services offered by a DBMS. It is therefore pertinent to identify key risks to the secure operation of a database system. This paper identifies the key risks, namely, SQL injection, weak audit trails, access management issues and issues with encryption. A malicious actor can get help from any of these issues. It can compromise integrity, availability and confidentiality of the data present in database systems. The paper also identifies various means and ways to defend against these issues and remedy them. This paper then proceeds to identify from the literature, the potential solutions to these ameliorate the threat from these vulnerabilities. It proposes the usage of encryption to protect the data from being breached and leveraging encrypted databases such as CryptoDB. Better access control norms are suggested to prevent unauthorized access, modification and deletion of the data. The paper also recommends ways to prevent SQL injection attacks through techniques such as prepared statements.

For multi-source heterogeneous complex data types of data cleaning and visual display, we proposed to build dynamic multimode visualization analysis tool, according to the different types of data designed by the user in accordance with the data model, and use visualization technology tools to build and use CQRS technology to design, external interface using a RESTFul architecture, The domain model and data query are completely separated, and the underlying data store adopts Hbase, ES and relational database. Drools is adopted in the data flow engine. According to the internal algorithm, three kinds of graphs can be output, namely, transaction relationship network analysis graph, capital flow analysis graph and transaction timing analysis graph, which can reduce the difficulty of analysis and help users to analyze data in a more friendly way

Terrorism, and radicalization are major economic, political, and social issues faced by the world in today's era. The challenges that governments and citizens face in combating terrorism are growing by the day. Artificial intelligence, including machine learning and deep learning, has shown promising results in predicting terrorist attacks. In this paper, we attempted to build a machine learning model to predict terror activities using a global terrorism database in both relational and graphical forms. Using the Neo4j Sandbox, you can create a graph database from a relational database. We used the node2vec algorithm from Neo4j Sandbox's graph data science library to convert the high-dimensional graph to a low-dimensional vector form. In order to predict terror activities, seven machine learning models were used, and the performance parameters that were calculated were accuracy, precision, recall, and F1 score. According to our findings, the Logistic Regression model was the best performing model which was able to classify the dataset with an accuracy of 0.90, recall of 0.94 precision of 0.93, and an F1 score of 0.93.

It is proposed to address existing methodological issues in the educational process with the development of intellectual technologies and knowledge representation systems to improve the efficiency of higher education institutions. For this purpose, the structure of relational database is proposed, it will store the information about defended dissertations in the form of a set of attributes (heuristics), representing the mandatory qualification attributes of theses. An inference algorithm is proposed to process the information. This algorithm represents an artificial intelligence, its work is aimed at generating queries based on the applicant preferences. The result of the algorithm's work will be a set of choices, presented in ranked order. Given technologies will allow applicants to quickly become familiar with known scientific results and serve as a starting point for new research. The demand for co-researcher practice in solving the problem of updating the projective thinking methodology and managing the scientific research process has been justified. This article pays attention to the existing parallels between the concepts of technical and human sciences in the framework of their convergence. The concepts of being (economic good and economic utility) and the concepts of consciousness (humanitarian economic good and humanitarian economic utility) are used to form projective thinking. They form direct and inverse correspondences of technology and humanitarian practice in the techno-humanitarian mathematical space. It is proposed to place processed information from the language of context-free formal grammar dissertation abstracts in this space. The principle of data manipulation based on formal languages with context-free grammar allows to create new structures of subject areas in terms of applicants' preferences.It is believed that the success of applicants’ work depends directly on the cognitive training of applicants, which needs to be practiced psychologically. This practice is based on deepening the objectivity and adequacy qualities of obtaining information on the basis of heuristic methods. It requires increased attention and development of intelligence. The paper studies the use of heuristic methods by applicants to find new research directions leads to several promising results. These results can be perceived as potential options in future research. This contributes to an increase in the level of retention of higher education professionals.

Managing and storing big data is non-trivial for traditional relational databases (RDBMS). Therefore, the NoSQL (Not Only SQL) database management system emerged. It is ca-pable of handling the vast amount and the heterogeneity of data. In this research, we are interested in one of its trending types, the graph database, namely, the Directed Property Graph (DPG). This type of database is powerful in dealing with complex relationships (\$\textbackslashmathrme.\textbackslashmathrmg\$., social networks). However, its sen-sitive and private data must be protected against unauthorized access. This research proposes a security model that aims at exploiting and combining the benefits of Access Control, View-Based, and Query-Rewriting approaches. This is a novel combination for securing DPG.

Nowadays, data is essential in several fields, such as science, finance, medicine, and transportation, which means its value continues to rise. Relational databases are vulnerable to copyright threats when transmitted and shared as a carrier of data. The watermarking technique is seen as a partial solution to the problem of securing copyright ownership. However, most of them are currently restricted to numerical attributes in relational databases, limiting their versatility. Furthermore, they modify the source data to a large extent, failing to keep the characteristics of the original database, and they are susceptible to solid malicious attacks. This paper proposes a new robust reversible watermarking technique, Fields Based Inserting Position Tuples algorithm (FBIPT), for relational databases. FBIPT does not modify the original database directly; instead, it inserts some position tuples based on three Fields―Group Field, Feature Field, and Control Field. Field information can be calculated by numeric attributes and any attribute that can be transformed into binary bits. FBIPT technique retains all the characteristics of the source database, and experimental results prove the effectiveness of FBIPT and show its highly robust performance compared to state-of-the-art watermarking schemes.

The outsourcing of databases is very popular among IT companies and industries. It acts as a solution for businesses to ensure availability of the data for their users. The solution of outsourcing the database is to encrypt the data in a form where the database service provider can perform relational operations over the encrypted database. At the same time, the associated security risk of data leakage prevents many potential industries from deploying it. In this paper, we present a secure outsourcing database search scheme (BASDB) with the use of a smart contract for search operation over index of encrypted database and storing encrypted relational database in the cloud. Our proposed scheme BASDB is a simple and practical solution for effective search on encrypted relations and is well resistant to information leakage against attacks like search and access pattern leakage.

In this paper, we proposed a data security model of a big data analytical environment in the financial sector. Big Data can be seen as a trend in the advancement of technology that has opened the door to a new approach to understanding and decision making that is used to describe the vast amount of data (structured, unstructured and semi-structured) that is too time consuming and costly to load a relational database for analysis. The increase in cybercriminal attacks on an organization’s assets results in organizations beginning to invest in and care more about their cybersecurity points and controls. The management of business-critical data is an important point for which robust cybersecurity controls should be considered. The proposed model is applied in a datalake and allows the identification of security gaps on an analytical repository, a cybersecurity risk analysis, design of security components and an assessment of inherent risks on high criticality data in a repository of a regulated financial institution. The proposal was validated in financial entities in Lima, Peru. Proofs of concept of the model were carried out to measure the level of maturity focused on: leadership and commitment, risk management, protection control, event detection and risk management. Preliminary results allowed placing the entities in level 3 of the model, knowing their greatest weaknesses, strengths and how these can affect the fulfillment of business objectives.