نشریه سیستم های پردازشی و ارتباطی چند رسانه ای هوشمند
سال چهارم شماره 1 (پیاپی 11، بهار 1402)

In recent decades, rapid development in the world of technology and networks has achieved, also there is a spread of Internet of thing services in all fields over the world. Piracy numbers have increased, also a lot of modern systems were penetrated. Thus the developing information security technologies to detect the new attack become an important requirement. One of the most important information security technologies is an Intrusion Detection System (IDS) that uses machine learning and deep learning techniques to detect anomalies in the network. In all of the information processing systems, detecting cyber-attacks is one of the main challenges and its effects can be blocked or limited by timely detection of attacks. The IoT system is no exception to this phenomenon, and with the high development of this technology and the expansion of its infrastructure, the need for an intelligent intrusion detection system with high accuracy and speed is essential. Neural networks are modern systems and computational methods for machine learning, knowledge representation, and the application of acquired knowledge to maximize the output accuracy of complex systems. Neural networks have already been used to solve many problems related to pattern recognition, data mining, data compression and research is still underway with regards to intrusion detection systems. One of the disadvantages of using training with classical methods in neural networks is getting stuck in local optimal points. In this paper, we use the meta-heuristic algorithm of Imperial competition algorithm (ICA) to train neural networks and show that in the field of intrusion detection in the IoT system, it can show much better accuracy and speed to classical training methods. Results show that our proposed method has 90% accuracy. This method has a better performance in comparison to classical neural network that has 75% accuracy. In this article, we will show that the use of imperial competition evolutionary optimization algorithms instead of traditional methods can increase the accuracy of the IDS system. In addition, evolutionary optimization algorithms are zero order and less complicated than gradient methods. Therefore, using this method, in addition to reducing the cost of system implementation, can increase the speed and accuracy of intrusion detection. In addition, from reliability point of view, we will show that the ICA-based systems are more stable in different implementations.

In recent years, convexity theory has experienced rapid development. Many researchers have expanded it. A notable generalization of the convex function is the Inox function introduced and studied by Hanson. This has greatly expanded the role of inox in optimization. Ben and Mond introduced a class of functions called generalizations of pre-Inox functions. Regarding change inequalities and related problems in recent years, Pite and Postloche introduced the concept of pseudo pre-invex, and applied it in theoretical mechanics and nonlinear optimization. Later, Pite and Antchak introduced this concept of invex, and applied it to vector optimization. This shows that pre-invexity plays an important role in the development of various fields of pure and applied sciences. In this article, we first introduce the concept of  -pre-inox and then state and prove the Hermit-Hadamard theorem for it, and we will state that the previously obtained inequalities are a direct result of our main theorem. In this article, we first introduce the concept of  invex sets and  pre-invex functions. Then state and prove the Hermit-Hadamard theorem for it, and state that the previous inequalities are a direct result of our main theorem. Quantum calculus is known as the study of differential and integral calculus without restrictions. Euler (1783-1707) was the first to study quantum calculus. He introduced q in Newton's infinite series compositions. In the early 20th century, the study of quantum calculus was started by Jackson. In quantum computing, we obtain mathematical equivalents of q objects that can be recovered as . Note that quantum calculus is a subset of time scale calculus. The time scale of calculus provides a unified framework for the study of dynamical equations in both discrete and continuous domains. That in quantum calculus, we deal with a specific time scale called the q time scale. Quantum calculus is a bridge between mathematics and physics. Due to the significant applications of quantum computing in mathematics and physics, this issue has been the focus of many researchers. As a result, quantum calculus has emerged as a fascinating field. In recent years, convexity theory has experienced rapid development. Many researchers have expanded it. A notable generalization of the convex function is the inex function, which was introduced and studied by Hanson [1]. This has greatly expanded the role of inox in optimization. Ben and Mond [2] introduced a class of functions called generalizations of pre-invex functions.This fundamental result of Hermit and Hadamard (HH) has obtained by many mathematicians, and as a result, this inequality has been extended by Noor in various ways using new ideas and obtained the Hermit-Hadamard inequality for pre-invex functions. Regarding variational inequalities and related problems in recent years, Pite and Postloche [3],[4]  and [5] introduced the concept of pseudo pre-invex, and applied it in theoretical mechanics and nonlinear optimization. Later, Pite and Antchak [6] introduced this concept of inexity, and applied it in vector optimization. This shows that pre-invex plays an important role in the development of various fields of pure and applied sciences. For more details on the quantum calculus see references [7].MethodNo method applicable. In This article, we improve the Hermite-Hadamard (HH) integral inequality for  - preinvex functions.Assuming , Theorem 3.1 of the article [8], that is, relation (3) is obtained. Assuming , the main theorem of the article [9] and  and , Hermit Hadamard's main inequality is obtained.

A social network is a social structure made up of individuals or organizations. Social network analysis is an approach in which the network is considered as a set of nodes and relationships between them. Nodes are individuals and actually actors in the network and the relationships between them are displayed as connections between nodes. Among many social network analysis issues, link prediction has attracted much attention due to the growing number of social network users. Link prediction means predicting which new interaction is going to happen in the future. Traditional link prediction methods considered pairs of nodes as a unit and made decisions based on commonalities between them. In addition, we proposed a new similarity measure for link prediction in social networks. We compared this criterion with four prediction methods of Jaccard link, Salton Index, Salton Cosine, and resource allocation). Experimental runs in this article were carried out on five social network datasets. Our results showed that this criterion performed better than other link prediction techniques on all datasets. Social network analysis has recently attracted lots of attention among researchers due to its wide applicability in capturing social interactions. Link prediction, related to the likelihood of having a link between two nodes of the network that are not connected, is a key problem in social network analysis. Many methods have been proposed to solve the problem. Among these methods, similarity-based methods exhibit good efficiency by considering the network structure and using as a fundamental criterion the number of common neighbors between two nodes to establish structural similarity.

Domain adaptation has become an important issue today. A high percentage of data processing domain adaptation is done with a significant percentage of studies related to deep learning. Traditional methods often ignore the distance between the intra-class in source domain and target domain. As a result, models can be sensitive to outliers and noisy data, additionally increasing the negative transfer in the model. This method applied GAN to extract appropriate features and then used Fuzzy c-means to cluster train datasets in the target domain. Finally, based on the WMMD metric and CNN, the model estimates the final label data. Five real datasets are selected to generate eight transfer tasks. The results show that the superiority of the proposed model lies in transferring more knowledge from the source domain to the target domain. In this approach, firstly based on GAN extracting features from source domains and the target domain (without labels), then label estimation by Fuzzy c-means clustering, finding the center of Fuzzy c-means on target domain data, new data points with labels in target domain as a new input to feature extraction module and regenerate features by GAN based on new pseudo labels. Afterward, we apply WMMD metrics based on CNN to ultimately assign labels for the target domain. Consequently, classification tasks have been done. Empirical results on various benchmark datasets showcase the exceptional performance of the proposed method compared to state-of-the-art DA approaches, validating the proposed Deep-Learning Unsupervised Domain Adaptation approach efficacy. Overall, the approach shows potential for advancing domain adaptation research by offering an efficient and resilient approach for addressing domain shifts in real-world applications. Experimental results on visual object recognition and a digit dataset reveal that the proposed algorithm is robust, flexible, and significantly superior regarding accuracy compared to the baseline DA approaches. Based on the three and combined digit datasets, 1.7% and 2.4% accuracy improvement are achieved, respectively, compared to the best baseline DA approach results. In this research, we addressed the challenging issues of outlier and negative transfer in the context of domain adaptation. Despite significant progress in domain adaptation techniques, outliers and negative transfer instances continue to hinder models' generalization performance across different domains. Based on DNNs and the WMMD metric, our proposed method was designed to mitigate these issues and effectively enhance knowledge transfer between domains.

This paper presents a novel scalable Constraint Search Based method forTest Data Generation, called CSBTDG, used in structural testing. CSBTDG outperforms existing methods for test data generation in several orders of magnitude, both in time and constraint efficiency. Search-based software testing is a powerful automated method to generate test inputs for software under test. Its goal is to reach a branch or a statement in a program. One major limitation of this approach is an insufficiently informed fitness function to start and guide search toward a test target within nested predicates (constraints). Another important limitation of this approach about equality constraints. To address these problems we propose a new search method based on integrating constraint programming using choco constraint solver and the Gray Wolf heuristic search algorithm. Preliminary experiments promise efficiency and effectiveness for the new constraint search based test data generation approach using multiple fitness functions.Test data generation is done by solving constraints related to test paths extracted from the program under test [3][2]. In general, the constraint solving problem of the test paths is a sub problem of a wider set of problems called constraint satisfaction problem (CSP) [4]. Formally, a constraint satisfaction problem is defined as a triple (X, D, C). X represents a set of variables, D is a set of range of values for variables and C is a set of constraints. Constraint satisfaction is the process of finding solutions for a set of constraints that satisfy the conditions imposed on the variables. Evaluation of variables is a function of a subset of variables to a specific set of values in the corresponding domains. The proposed solution is called consistent evaluation if none of the constraints are violated. An evaluation is a complete evaluation if it includes all variables. If the assessment is consistent and complete, it is a solution. It is said that such evaluation is a solution to the constraint satisfaction problem. Therefore, a solution is a set of values for the variables that satisfy all the constraints, that is, a point in the feasible region. The main innovation of this article is to present a cooperation model to exchange information between the solver and the searcher. In this model of data generation, the test solver is an initial setup to assign values to some variables to start searching for the meta-heuristic algorithm.

The widespread use of the internet and social media platforms has led to an explosion of digital data, including users' opinions about various services and products. These opinions are valuable sources of information for businesses and organizations to understand the needs and preferences of their customers. Supervised machine learning models have been proven to be effective in analyzing users' opinions. However, to achieve efficient results, a sufficient amount of labeled training data is necessary. Labeling data requires a considerable amount of time and resources, which can be a significant challenge for many organizations. This is where the concept of semi-supervised learning comes in, which utilizes both labeled and unlabeled data to improve the performance of the model.
In this paper, a semi-supervised approach to analyze users' Persian opinions has been proposed. The method takes advantage of the abundant unlabeled data available in addition to a small number of labeled data in the training phase. The proposed method uses the support vector machine (SVM) algorithm, which has been shown to be effective in opinion mining in related research. The proposed method extracts emotional words from comments using sentiment lexicons and then extracts term frequency-inverse of document frequency vectors. The semi-supervised SVM algorithm is then applied to these vectors to estimate the polarity of sentiments.
To evaluate the performance of the proposed method, it has been tested on the Digikala comments dataset and compared with the supervised SVM algorithm and semi-supervised self-training method for different numbers of labeled data based on accuracy, precision, recall, and F1 criteria. The results indicate that the proposed semi-supervised method outperforms the supervised SVM algorithm and the semi-supervised method of self-training. The impact of the size of unlabeled data is also investigated in the experiments.
One of the advantages of the proposed method is that it can estimate the polarity of opinions that have not been trained in the training phase, which is not possible in some graph-based methods. Furthermore, it is not affected by the error of training with labeled data in self-training methods. In conclusion, the proposed semi-supervised method provides an efficient solution for analyzing users' opinions in Persian. This method can be used by businesses and organizations to gain insights into their customers' opinions and improve their products and services accordingly.