نشریه فناوری اطلاعات و ارتباطات انتظامی
پیاپی 13 (بهار 1402)

The image fusion process is the process of combining two or more images with the aim of achieving a desired image. The image obtained during this process contains more information and better accuracy than each of the input images. Medical images and satellite SAR images generally have pulse noises. These noises are sometimes caused by the shift register error and during data transmission. The existence of this noise is one of the challenges of using these images for fusion, in such a way that their fusion with the usual methods does not work well. Therefore, in this article, a hybrid approach based on variable mode analysis is proposed with the aim of eliminating multiplicative noise during image fusion. Pulse noise is common in medical ultrasound images. Also, this noise is present in the images of SAR radars, which are also used in environmental monitoring such as border patrol or maritime patrol due to their characteristics. The proposed solution has been simulated and implemented in the MATLAB software environment. Also, its efficiency has been evaluated in two indices of noise-to-signal ratio and peak noise-to-signal ratio. The proposed method of the article improved the SNR from 7.7416 to 12.4895 and the PSNR from 16.7294 to 20.3549 compared to the VMD method.

At the international level and in Iran, social networks and sites focus more on the quality of their products according to the opinions of their users and do not pay attention to the level of security and reliability, and even when the behavior of customers is analyzed, No attention is paid to unusual and profit-seeking users. The main idea of this research is to investigate the behavior of customers and predict their behavior in the future based on deep learning in order to increase the quality of service to valid users. In this research, a method based on deep learning technique is presented, with the help of which the behavior of users in online stores is analyzed. In the proposed method of this research, the deep neural network algorithm is used to train a set of data after the initial processing of user behavior and feature extraction. The proposed method has been implemented in MATLAB software. To evaluate the method of this research, we have used conventional criteria that are used in data mining applications, such as accuracy, readability and F score. According to the experiments, it can be seen that the proposed method is the first priority in terms of data mining evaluation criteria. By applying the methods and using the results of this research, it is possible to predict the behavior of customers on certain sites in order to create security and reduce risks for various activities in the information society and protect the values of the society.

A smart contract is a computer protocol for creating an agreement, which plays an important role in many fields. The terms of the contract are coded after meeting the conditions, automatically and in the platform of the blockchain. Developing these types of contracts, despite being called smart, is a tedious task beyond defining a set of rules. To program smart contract code, in addition to the need to have expertise in a specific domain, the need to have strong technical knowledge in the field of smart contract programming languages and new blockchain platforms is essential. One of the most common smart contract languages is the Solidity language based on the Ethereum blockchain. To reduce the need for technical knowledge related to smart contract programming languages, one of the solutions is to use modeling principles in the "Model-Driven Engineering" approach. Development using modeling in Model-Driven Engineering can reduce or eliminate the effort required for coding. In this paper, we present a framework based on model-driven architecture called SCoMoLa for developing Solidity smart contracts. In order to evaluate it, three criteria of maintainability, understandability, and extensibility have been studied on five smart contract samples. Furthermore, based on these criteria, the proposed metamodels are compared with the metamodels of two other Model-Driven approaches, and the superiority of the proposed solution is demonstrated.

Radio frequency oscillators are used in all modern wireless telecommunications, radar, and remote sensing systems. In this article, a microwave oscillator based on a microstrip dual-band bandpass filter, with a central frequency of 1.95 and 5.45 GHz, has been designed and simulated for use in local wireless networks. The oscillation frequencies of the designed oscillator are asynchronous. The dual band pass band filter is designed as a feedback block of the oscillator in the direction of frequency stability and also as a frequency selector element. Some filter parameters such as quality factor, transmission loss, number of transmission zeros, bandwidth, and return loss are effective in the phase noise response of the oscillator. In this way, the phase noise of the microwave oscillator with feedback has a direct relationship with the quality coefficient of the resonators used in its structure, and with the feedback loss, which is the insertion loss of the filter. All these parameters in order to reduce the phase noise, the selection and the oscillator oscillate at the same peak of the quality coefficient of the intermediate filter to reduce the phase noise. In this design, the phase noise is less than -150 dBc/Hz at the offset frequency of 1 MHz due to the application of the oscillator in local wireless networks. All stages of design and simulation have been implemented using ADS software.

Today, due to the spread of social networks and increasing the number of users, the use of information in these networks has become one of the most important issues in various fields. Using natural language processing techniques for political analysis can be one of the effective ways to improve the goals of different societies in the field of politics.In this research to predict the users’ political tendency, a proposed deep learning model and Twitter’s text data are used. The aim is to present a suitable model with high accuracy to predict the vote results in English and Persian languages. The proposed approach has been tested on the dataset collected in Persian and English from the Twitter social network and the results obtained have been compared with the results of the existing methods.The proposed model consists of a deep convolutional neural network with long short-term memory (CNN-BiLSTM) in such a way that it simultaneously learns the semantic relation of the text and uses the advantages of CNN and BiLSTM deep networks simultaneously. The proposed algorithm has achieved 84.5% accuracy for Persian data and 96.18% accuracy for English data.

The use of artificial intelligence for the scientific detection of crimes is a requirement in smart police, and the analysis of video images to detect target events using neural networks is a general goal in this article. Video recovery systems with content-based recovery method is a suitable approach. The distinctive feature of this method is the recovery of visual and perceptual contents hidden in the video, which is the same mechanism by which humans can understand the difference between videos. With the increase in people's use of video in all aspects of daily life, such as news, sports competitions, etc., the need to search and find video has also expanded. Considering that video can contain a lot of content in a short time, the best search method that can extract the most similar videos is video query. . Using the features and general information of the frames may cause the loss of the content and the main topic of the videos. Therefore, in this article, the role of the object of interest in the video has been investigated. Feature extraction from frames has been done using VGG-16 neural network. Also, a rating criterion based on the similarity of frames, the sequence of similar frames and the number of frames similar to the searched video has been introduced. The efficiency of the proposed method has been checked on the YouTube action dataset with the average accuracy criterion. The obtained results show 94.90% accuracy in retrieving the top ten samples.

The complexity and inconsistency in the environment is increasing strongly, and smart organizations need to adapt to the environment in order to perform smart and timely actions, in other words, acting smart is an integral part of increasing their competitiveness and survival. In order to achieve goals and missions in such organizations, information security is of high importance. In order to deal with advanced threats, appropriate solutions are needed for prevention, detection and response in an effective time frame. Meanwhile, the organization of knowledge and the use of common concepts by humans and computer agents has become an important issue in the field of security, and ontology is presented as the key to this issue. Considering the nature and capabilities hidden in ontology, new and innovative applications are envisioned in the field of intelligent organization security, which has made it a suitable tool for security knowledge modeling. In this research, by studying the world's successful achievements and experiences in this field and research in the field of strategic documents in terms of requirements, the network ontology design was done in the Protégé software environment and the model was presented using Onto Graf, and then the application software as The semantic infrastructure of the network was designed with Python language.Therefore, it is possible to continuously develop the ontological model and provide a common understanding in the organization, and based on the conclusions obtained from the model and the designed software, provide suitable solutions to increase the security of the intelligent organization

With the development of information technology (IT) industries, information security has become one of the critical issues. Today, blockchain technology is widely used to improve information security. Most organizations use centralized databases to manage systems. In this method, saving and retrieving information is done at one point and the records are updated after confirmation. Non-transparency, changeability and involvement of the central manager in verifying the correctness of the data are among the disadvantages of centralized methods. But blockchain is a tool for individual and organization to exchange digital assets without the intervention of a trusted third party, i.e. a central administrator. Blockchain has several key features, such as scalability, immutability and transparency. The purpose of this study is to evaluate the HyperledgerFabric platform in the block chain authorized for organizational use such as security institutions and the police, which deals with the research of the HyperledgerFabric platform and its applications in different fields of research, such as smart police, by stating the types of block chain in terms of the type of implementation. to be Finally, the path of future actions and open research challenges in this field are also discussed in detail, and by introducing examples of research conducted in the last three years, achievements, created products, applications and methods of implementation, and expressing strengths and weaknesses. introduces the HyperledgerFabric platform as a block chain suitable for developing the services of organizations, especially security and police institutions, through the distribution of databases on the block chain platform.

In this paper, a new histogram-based method is introduced to make object detectors resistant to hostile attacks. In the following, this method was applied to two object detector models, YOLOV5 and FRCNN, and in this way, two models resistant to attacks were introduced. In order to verify the performance of the mentioned models, we performed the adversarial training process of these models with three targeted attacks TOG-vanishing, TOG-mislabeling, and TOG-fabrication and one untargeted attack, DAG. We have checked the efficiency of the introduced models on two data sets MSCOCO and PASCAL VOC, which are among the most famous data sets in the field of object recognition. The results show that this method, in addition to improving the adversarial accuracy, also improves the clean accuracy of the object detector models to some extent. The average clean accuracy of the YOLOv5-n model for the PASCAL VOC dataset, if adversarial attacks are applied to it, in the case where no defense method is applied, is 85.5%, and in the case where the histogram method is applied, the average accuracy is equal to with 87.36%. In the YOLOv5-n model, according to the results, the best adversarial accuracy of this model, which has increased compared to other models, is in TOG-vanishing and TOG-fabrication attacks, which are 48% and 52.36%, respectively.

Secret Sharing is an important branch of cryptography that is used to protect documents and information from risks such as theft and unauthorized access. In Secret Sharing schemes, shareholders offer a shadow share of the image to enhance security when recovering secret images, rather than the original share. Thus, each shareholder with only one share of the secret image is able to share many images of the secret with others. A new (k,n) threshold secret image sharing scheme base on finite geometry is introduced in paper. In the proposed scheme the secret image is first partitioned into several non-overlapping blocks of k pixels. Every set of k pixel forms a point under a k-dimensional space and the set solution to each generated affine hyperplane, only intersecting at this point stores to the corresponding shared image. The original secret image can be reconstructed by gathering any k or more encrypted shared images. The experimental results show that the proposed image secret sharing scheme is an efficient and secure method.