Anomaly Detection (AD) and Anomaly Localization (AL) are crucial in fields that demand high reliability, such as medical imaging and industrial monitoring. However, current AD and AL approaches are often susceptible to adversarial attacks due to limitations in training data, which typically include only normal, unlabeled samples. This study introduces PatchGuard, an adversarially robust AD and AL method that incorporates pseudo anomalies with localization masks within a Vision Transformer (ViT)-based architecture to address these vulnerabilities. We begin by examining the essential properties of pseudo anomalies, and follow it by providing theoretical insights into the attention mechanisms required to enhance the adversarial robustness of AD and AL systems. We then present our approach, which leverages Foreground-Aware Pseudo-Anomalies to overcome the deficiencies of previous anomaly-aware methods. Our method incorporates these crafted pseudo-anomaly samples into a ViT-based framework, with adversarial training guided by a novel loss function designed to improve model robustness, as supported by our theoretical analysis. Experimental results on well-established industrial and medical datasets demonstrate that PatchGuard significantly outperforms previous methods in adversarial settings, achieving performance gains of 53.2% in AD and 68.5% in AL, while also maintaining competitive accuracy in non-adversarial settings. The code repository is available at here.

Despite significant progress in Anomaly Detection (AD), the robustness of existing detection methods against adversarial attacks remains a challenge, compromising their reliability in critical real-world applications such as autonomous driving. This issue primarily arises from the AD setup, which assumes that training data is limited to a group of unlabeled normal samples, making the detectors vulnerable to adversarial anomaly samples during testing. Additionally, implementing adversarial training as a safeguard encounters difficulties, such as formulating an effective objective function without access to labels. An ideal objective function for adversarial training in AD should promote strong perturbations both within and between the normal and anomaly groups to maximize margin between normal and anomaly distribution. To address these issues, we first propose crafting a pseudo-anomaly group derived from normal group samples. Then, we demonstrate that adversarial training with contrastive loss could serve as an ideal objective function, as it creates both inter- and intra-group perturbations. However, we notice that spurious negative pairs compromise the conventional contrastive loss for achieving robust AD. Spurious negative pairs are those that should be mapped closely but are erroneously separated. These pairs introduce noise and misguide the direction of inter-group adversarial perturbations. To overcome the effect of spurious negative pairs, we define opposite pairs and adversarially pull them apart to strengthen inter-group perturbations. Experimental results demonstrate our superior performance in both clean and adversarial scenarios, with a 26.1% improvement in robust detection across various challenging benchmark datasets.

Novelty detection is a critical task for deploying machine learning models in the open world. A crucial property of novelty detection methods is universality, which can be interpreted as generalization across various distributions of training or test data. More precisely, for novelty detection, distribution shifts may occur in the training set or the test set. Shifts in the training set refer to cases where we train a novelty detector on a new dataset and expect strong transferability. Conversely, distribution shifts in the test set indicate the methods' performance when the trained model encounters a shifted test sample. We experimentally show that existing methods falter in maintaining universality, which stems from their rigid inductive biases. Motivated by this, we aim for more generalized techniques that have more adaptable inductive biases. In this context, we leverage the fact that contrastive learning provides an efficient framework to easily switch and adapt to new inductive biases through the proper choice of augmentations in forming the negative pairs. We propose a novel probabilistic auto-negative pair generation method (AutoAugOOD), along with contrastive learning, to yield a universal novelty detector method. Our experiments demonstrate the superiority of our method under different distribution shifts in various image benchmark datasets. Notably, our method emerges universality in the lens of adaptability to different setups of novelty detection, including one-class, unlabeled multi-class, and labeled multi-class settings.

Scanning for trojan (backdoor) in deep neural networks is crucial due to their significant real-world applications. There has been an increasing focus on developing effective general trojan scanning methods across various trojan attacks. Despite advancements, there remains a shortage of methods that perform effectively without preconceived assumptions about the backdoor attack method. Additionally, we have observed that current methods struggle to identify classifiers trojaned using adversarial training. Motivated by these challenges, our study introduces a novel scanning method named TRODO (TROjan scanning by Detection of adversarial shifts in Out-of-distribution samples). TRODO leverages the concept of "blind spots"—regions where trojaned classifiers erroneously identify out-of-distribution (OOD) samples as in-distribution (ID). We scan for these blind spots by adversarially shifting OOD samples towards in-distribution. The increased likelihood of perturbed OOD samples being classified as ID serves as a signature for trojan detection. TRODO is both trojan and label mapping agnostic, effective even against adversarially trained trojaned classifiers. It is applicable even in scenarios where training data is absent, demonstrating high accuracy and adaptability across various scenarios and datasets, highlighting its potential as a robust trojan scanning strategy.

Abstract We propose a novel hybrid learning approach to gain situation awareness in smart environments by introducing a new situation identifier that combines an expert system and a machine learning approach. Traditionally, expert systems and machine learning approaches have been widely used independently to detect ongoing situations as the main functionality in smart environments in various domains. Expert systems lack the functionality to adapt the system to each user and are expensive to design based on each setting. On the other hand, machine learning approaches fail in the challenge of cold start and making explainable decisions. Using both of these approaches enables the system to use user’s feedback and capture environmental changes while exploiting the initial expert knowledge to solve the mentioned challenges. We use decision trees and situation templates as the core structure to interpret sensor data. To evaluate the proposed method, we generate a new human-annotated dataset simulating a smart environment. Our experiments show superior results compared with the initial expert system and the machine learning approach while preserving the initial expert system’s interpretability.

With the widespread access of people to the Internet and the increasing usage of social networks in all nations, social networks have become a new source to study cultural similarities and differences. We identified major issues in traditional methods of data collection in cross-cultural studies: difficulty in access to people from many nations, limited number of samples, negative effects of translation, positive self-enhancement illusion, and a few unreported problems. These issues are either causing difficulty to perform a cross-cultural study or have negative impacts on the validity of the final results. In this paper, we propose a framework that aims to calculate cultural distance among several countries using the information and cultural features extracted from social networks. To this aim, the framework estimates the distribution of news-oriented tweets for each nation and computes the cultural distance from these sets of distributions. Based on a sample composed of more than 17 million tweets from late 2017, our framework calculated cultural distance between 22 countries. Our results show a positive correlation between cultural distances computed by our framework and distances computed by Hofstede's cultural scores and also identified connections between some of the cultural features.

Social norms are a core concept in social sciences and play a critical role in regulating a society's behavior. Organizations and even governmental bodies use this social component to tackle varying challenges in the society, as it is a less costly alternative to establishing new laws and regulations. Social networks are an important and effective infrastructure in which social norms can evolve. Therefore, there is a need for theoretical models for studying the spread of social norms in social networks. In this paper, by using the intrinsic properties of norms, we redefine and tune the Rescorla-Wagner conditioning model in order to obtain an effective model for the spread of social norms. We extend this model for a network of people as a Markov chain. The potential structures of steady states in this process are studied. Then, we formulate the problem of maximizing the adoption of social norms in a social network by finding the best set of initial norm adopters. Finally, we propose an algorithm for solving this problem that runs in polynomial time and experiments it on different networks. Our experiments show that our algorithm has superior performance over other methods.

In Software Product Line (SPL), Configuration Management (CM) is a multi-dimensional problem. On the one hand, the Core Assets that constitute a configuration need to be managed, and on the other hand, each product in the product line that is built using a configuration must be managed, and furthermore, the management of all these configurations must be coordinated under a single process. Therefore, CM for product lines is more complex than for single systems. The CM of any software system involves four closely related activities: Change Management (ChM), Version Management (VM), System Building (SB) and Release Management (RM) [I. Sommerville, Software Engineering, 9th edn. (Addison-Wesley, 2010)]. The aim of this paper is to provide ChM and VM models for evolutionary-based SPL system development and maintenance. The proposed models support any level of aggregation in SPLs and have been applied to Mobile SPL as a case study.

The degree distribution is an important characteristic of complex networks. In many data analysis applications, the networks should be represented as fixed-length feature vectors and therefore the feature extraction from the degree distribution is a necessary step. Moreover, many applications need a similarity function for comparison of complex networks based on their degree distributions. Such a similarity measure has many applications, including classification and clustering of network instances, evaluation of network sampling methods, anomaly detection and study of epidemic dynamics. The existing methods are unable to effectively capture the similarity of degree distributions, particularly when the corresponding networks have different sizes. In this paper, we propose a feature extraction method and a similarity function for the degree distributions in complex networks. We propose to calculate the feature values based on the mean and standard deviation of the node degrees in order to decrease the effect of the network size on the extracted features. Experiments on a wide range of real and artificial networks confirms the accuracy, stability and effectiveness of the proposed method.

We propose a self-adaptive process (SAP) that maintains the software architecture quality using the MAPE-K standard model. The proposed process can be plugged into various software development processes and service-oriented methodologies due to its explicitly defined inputs and outputs. To this aim, the proposed SAP is integrated with the service-oriented modeling and application (SOMA) methodology in a two-layered structure to create a novel methodology, named self-adaptive service-oriented architecture methodology (SASOAM), which provides a semi-automatic self-aware method by the composition of architectural tactics. Moreover, the maintenance activity of SOMA is improved using architectural and adaptive patterns, which results in controlling the software architecture quality. The improvement in the maintainability of SOMA is demonstrated by an analytic hierarchy process (AHP) based evaluation method. Furthermore, the proposed method is applied to a case study to represent the feasibility and practicality of SASOAM.

In many popular applications like peer-to-peer systems, large amounts of data are distributed among multiple sources. Analysis of this data and identifying clusters is challenging due to processing, storage, and transmission costs. In this paper, we propose GDCluster, a general fully decentralized clustering method, which is capable of clustering dynamic and distributed data sets. Nodes continuously cooperate through decentralized gossip-based communication to maintain summarized views of the data set. We customize GDCluster for execution of the partition-based and density-based clustering methods on the summarized views, and also offer enhancements to the basic algorithm. Coping with dynamic data is made possible by gradually adapting the clustering model. Our experimental evaluations show that GDCluster can discover the clusters efficiently with scalable transmission cost, and also expose its supremacy in comparison to the popular method LSP2P.

Modeling is one of the major research areas in social network analysis whose goal is to study networks structure and its evolution. Motivated by the intuition that members in social networks behave selfishly, network creation games have been introduced for modeling social networks. In this paper, our aim is to measure how much the output graphs of a given network creation game are compatible with a social network. We first show that the precise measurement is not possible in polynomial time. Then we propose a method for its approximation; finally, we show the usability of our method by conducting experiments on real network data.

One of the arising issues in wireless sensor networks is running multiple applications to reduce deployment, administrative and management costs. In this paper we assume a sensor network in which nodes of a group have a set of same applications in their external flash memory that some of them are loaded and running. Occasionally, these applications should be updated for reasons that are manifold ranging from fixing software bugs to retasking. It is often that updates for a subset of applications happen at the same time, for example when a routing protocol shared among a subset of applications is changed. To bring the sensor nodes up to date with the new versions of the applications, existing strategies transfer the delta code, a smaller code obtained from the differential compression between the binary representations of the old and the new version of each application, and let the sensor nodes rebuild the new version from the received delta and the old version. We discuss that these strategies are blind to the order of updates when a set of updates occurs simultaneously. We introduce a heuristic planner that yields a sequence for a set of simultaneous updates, which minimizes the total delta size without adding overhead on the sensor nodes. We illustrate the efficiency of our planner on some TinyOS applications.

Nowadays, software product line is an approach to reduce costs of software development, decrease time to market, and increase capabilities of reuse in designing and exploiting software development processes. Moreover, other quality attributes of the project domain should be considered to enhance quality of the product. Meanwhile, taking advantage of software product line makes developers capable of estimating development costs and time to market in a more realistic way. However, old approaches to estimate cost of development and foresee time to market are not suitable enough for software product line. In this paper, some important business parameters and a way to calculate cost and time to market in a product line are presented. Changing components among time, portion of the change in a specific product and organization issues are observed in the estimation function.

Evaluating architectures and choosing the correct one is a critical issue in software engineering domain, in accordance with extremely extension of architecture-driven designs. In the first years of defining architecture styles, some special quality attributes were introduced as their basic attributes. After a moment, by utilizing them in practice, some results were obtained confirming some of attributes; some others meanwhile were not witnessed. As software architecture construction process is dependent on and addressed by both usage conditions and quality attributes, in this paper a framework has been proposed to provide an environment and a platform that can cover evaluation of architecture styles with a technique that not only exploits both qualitative and quantitative information but also considering users' needs is possible precisely and with high quality. Moreover, we define a classification and notation in order to describe heterogeneous architectures. It provides us with the ability of evaluating heterogeneous architecture styles of a software system.

This paper presents a new interpretable fuzzy classification system. Simulated annealing heuristic is employed to effectively investigate the large search space usually associated with classification problem. Here, two criteria are used to evaluate the proposed method. The first criterion is accuracy of extracted fuzzy if-then rules, and the other is comprehensibility of obtained rules. Experiments are performed with some data sets from UCI machine learning repository. Results are compared with several well-known classification algorithms, and show that the proposed approach provides more accurate and interpretable classification system.

Clustering is the problem of finding relations in a data set in an supervised manner. These relations can be extracted using the density of a data set, where density of a data point is defined as the number of data points around it. To find the number of data points around another point, region queries are adopted. Region queries are the most expensive construct in density based algorithm, so it should be optimized to enhance the performance of density based clustering algorithms specially on large data sets. Finding the optimum set of region queries to cover all the data points has been proven to be NP-complete. This optimum set is called the skeletal points of a data set. In this paper, we proposed a generic algorithms which fires region queries at most 6 times the optimum number of region queries (has 6 as approximation factor). Also, we have extend this generic algorithm to create a DBSCAN (the most wellknown density based algorithm) derivative, named ADBSCAN. Presented experimental results show that ADBSCAN has a better approximation to DBSCAN than the DBRS-H (the most well-known randomized density based algorithm) in terms of performance and quality of clustering, specially for large data sets.

Predicting latency between nodes on the Internet can have a significant impact on the performance of many services that use latency distances among nodes as a decision making input. Coordinate-based approaches are among the family of latency prediction techniques where latency between each pair of nodes is modeled as the virtual distance among those nodes over a virtual system. This article proposes the decentralized coordinate system (DCS) as a fully distributed system that does not rely on any infrastructure support from the underlying network. DCS uses a two-phase algorithm: first, each host is assigned rough coordinates, then, the rough estimation is refined in a way that it gradually converges to the accurate relative position of nodes. Simulation results demonstrate that the accuracy of DCS is competitive with existing network coordinate systems and that it can be considered as a good alternative for some applications.

With the advent of online social networks, the trust-based approach to recommendation has emerged which exploits the trust network among users and makes recommendations based on the ratings of trusted users in the network. In this paper, we introduce a two dimensional trust model which dynamically gets updated based on users's feedbacks, in contrast to static trust values in current trust models. Explorability measures the extent to which a user can rely on recommendations returned by the social network of a trusted user. Dependability represents the extent to which a user's own ratings can be trusted by users trusting him directly and indirectly. We propose a method to learn the values of explorability and dependability from raw trust data and feedback expressed by users on the recommendations they receive. Positive feedback will increase the trust and negative feedback will decrease the trust among users. We performed an evaluation on the Epinions dataset, demonstrating that exploiting user feedback results in lower prediction error compared to existing trust-based and collaborative filtering approaches.

Modeling languages are needed to describe the conceptual construct underlying software. Several modeling languages have been proposed during the last decades. Cognitive complexity is one of the common problems in designing modeling languages. Users have to split their attention and cognitive resources between two different tasks when working with complex language: solving the problem and understanding the elements composing the language. Several researches have been accomplished to evaluate cognitive complexity of modeling languages, among them, metric based and empirical approaches aremore important and convenient than others. In this paper, we compared these two methods. Results show that there is no significant relation between outputs generated by these approaches.

Haplotype information has become increasingly important in analyzing fine-scale molecular genetics data, Due to the mutated form in human genome; SNPs (Single Nucleotide Polymorphism) are responsible for some genetic diseases. As a consequence, obtaining all SNPs from human populations is one of the primary goals of studies in human genomics. In this paper, a data fusion method based on multiple parallel classifiers for reconstruction of haplotypes from a given sample Single Nucleotide Polymorphism (SNP) is proposed. First, we design a single greedy algorithm for solving haplotype reconstructions. [2] is used as an efficient approach to be combined with first classification method. The methods and information fusion approach are aimed specifically for increasing reconstruction rate of the problem in Minimum Error Correction Model (MEC) which is one of haplotyping problem models belonging to NP-hard class. Designing a parallel classifier, which helps us cover the single classifier’s weaknesses, was the focus of our research.

New advances in mobile computer technology, along with rapid growth of quality and quantity of wireless networks, introduce new capabilities, applications, and concerns to computer science and industry. Mobile payment, in which mobile customers use their phones for performing financial transactions, is a promising and exciting domain that has been rapidly developing recently. Although different mobile payment schemes has been designed and deployed worldwide, there is still a lack of research in achieving efficient and global mobile payment systems. Different business players and various technologies make the situation more complex. In this paper, we present MPaySmart, a secure and user-friendly mobile payment system which focuses on customer requirements. It is an operator-driven SIM-based mobile payment scheme, which utilizes Over-the-Air technology for secure communications.

Due to the enlargement and complexity of software systems and the need for maintenance and update, success of systems depends strongly on their architecture. Software architecture has been a key element in software development process in two past decades. Therefore, choosing the correct architecture is a critical issue in software engineering domain, with respect to the extremely extension of architecture-driven designs. Moreover, software architecture selection is a multi-criteria decision-making problem in which different goals and objectives should be considered. In this paper, a decision support system (DSS) has been designed which provides software architects with more precise and suitable decisions in architecture styles selection. The DSS uses fuzzy inference to support decisions of software architects and exploits properties of styles in the best way while making decisions.

Today's software systems are more frequently composed from preexisting commercial or non-commercial components and connectors. These components provide complex and independent functionality and are engaged in complex interactions. Component-Based Software Engineering (CBSE) is concerned with composing, selecting and designing such components. As the popularity of this approach and hence number of commercially available software components grows, selecting a set of components to satisfy a set of requirements while minimizing cost is becoming more difficult. This problem necessitates the design of efficient algorithms to automate component selection for software developing organizations. We address this challenge through analysis of Component Selection, the NP-complete process of selecting a minimal cost set of components to satisfy a set of objectives. Due to the high order of computational complexity of this problem, we examine approximating solutions that make the component selection process practicable. We adapt a greedy approach and a genetic algorithm to approximate this problem. We examined the performance of studied algorithms on a set of selected ActiveX components. Comparing the results of these two algorithms with the choices made by a group of human experts shows that we obtain better results using these approximation algorithms.

This paper presents a novel hybrid approach for intrusion detection in computer networks. The proposed approach combines an evolutionary based fuzzy system with an artificial immune system to generate high quality fuzzy classification rules. The performance of final fuzzy classification system has been investigated using the KDD-Cup99 benchmark dataset. The results indicate that in comparison to several traditional techniques, such as C4.5, Naive Bayes, k-NN and SVM, the proposed hybrid approach achieves better classification accuracies for most of the classes of the intrusion detection classification problem. Therefore, the resulted fuzzy classification rules can be used to produce a reliable intrusion detection system.

The process of monitoring the events occurring in a computer system or network and analyzing them for sign of intrusions is known as intrusion detection system (IDS). The objective of this paper is to extract fuzzy classification rules for intrusion detection in computer networks. The proposed method is based on the iterative rule learning approach (IRL) to fuzzy rule base system design. The fuzzy rule base is generated in an incremental fashion, in that the evolutionary algorithm optimizes one fuzzy classifier rule at a time. The performance of final fuzzy classification system has been investigated using intrusion detection problem as a high-dimensional classification problem. Results show that the presented algorithm produces fuzzy rules, which can be used to construct a reliable intrusion detection system.

In the last two decades, a number of architecture- based software development notations, methods, techniques, and tools were formulated. Of particular interest to the software architecture researchers and practitioners were the notations and languages for modeling software architectures, known as ADLs. Object stochastic activity network (OSAN) is a new model which inherits useful features from Petri nets, stochastic activity networks and queue networks, which made it a powerful and flexible tool to model concurrent systems and evaluate and verify them. This paper introduces OSAN as a new analyzable architecture description language, which can be used both for modeling architectural aspects of software systems and evaluating and analyzing them.

Abstract In overlay networks, the mechanism of a peer randomly joining and leaving a network, causes a topology mismatch between the overlay and the underlying physical topology. This causes a large volume of redundant traffic in the underlying physical network as well as an extra delay in message delivery in the overlay network. Topology mismatch occurs because overlay networks are not aware of their underlying physical networks. In this paper we present a mathematical model for topology awareness of overlay networks (degree of matching between an overlay and its underlying physical network) and the efficiency of message delivery on them. Then we experimentally show that the model is precise in the real world. Also, after determining the computational complexity of the model, we propose an optimization heuristic algorithm to enhance efficiency of message exchange on overlay networks by increasing their topology awareness. Next, we present the results of running the algorithm on different kinds of random graph and show how we can implement the algorithm on peer‐to‐peer networks; we then simulate the implementation on some unstructured peer‐to‐peer overlay networks. Copyright © 2006 John Wiley & Sons, Ltd.

One of the main requisites for multi agent researchers is a multi-purpose infrastructure that enables them to create, test and evaluate protocols, methods and algorithms. RoboCupRescue simulation environment is a flexible and configurable environment that can fill this gap. With changes in its configuration files, it will gain different features, so different methods can be designed, implemented, and tested. This environment is actually suitable for most multi-agent research issues such as cooperation, coordination, agent communication languages, and agent micro level issues. Sensible roles and tasks in the system motivate more research and make this environment a suitable infrastructure for educational purposes. Arian is a set of implemented agents for RoboCupRescue system. Using Arian agents will speed up implementing agents for this environment. In this paper, we introduce RoboCupRescue simulation environment and Arian agents along with some examples of research issues in this environment. We also present some of RoboCupRescue sub-problems for specific research goals. Although RoboCupRescue can be considered as an application of multi-agent research, in this paper we focus on the suitability of this environment as an infrastructure for multi-agent research and education.