American Research Journal of Computer Science and Information Technology

Oleksandr Khodorkovskyi 

The article discusses issues related to the specifics of using trained artificial intelligence (AI) to automate decision-making (with an emphasis on the industrial sector). The importance of addressing the topic is predetermined by the growing demands on the efficiency and reliability of industrial systems (especially against the background of rapid digitalization, which permeates all areas of activity without exception). In the current environment, trained AI is becoming a key tool to help establish and optimize production processes, as well as reduce operating costs. At the same time, its application in the field under consideration is accompanied by several contradictions, including the implied lack of unified approaches to its integration, limited methods for assessing the quality of those decisions that are made, and insufficient attention to issues of ethical responsibility. The purpose of the study is to analyze existing developments regarding the integration of trained AI into industrial systems, systematization (based on familiarization with scientific publications) of advantages, and related limitations. The article summarizes modern achievements in the field of predictive analytics, security management methods, and the protection of AI systems from adversarial attacks. The conclusions point to the need to unify methodological approaches, increase the resilience of AI to external threats, and develop standards governing its use. The presented materials will be useful to scientists dealing with the problems of artificial intelligence, and automation, specialists in industrial safety, and heads of enterprises interested in the digitalization of production.

Vladislav Terekhov 

Optimizing the power consumption of Android applications developed on the Java platform is a key factor in enhancing the user experience and extending the life of mobile devices. The high level of application power consumption is due to many factors, including working with sensors, network and processor. The article discusses the main optimization techniques, such as managing background processes, using data caching, reducing sensor and CPU activity, as well as using more efficient APIs and technologies such as WorkManager and App Standby Buckets. Code-level optimization and the competent use of third-party profiling tools can significantly reduce power consumption, extending the operating time of the device without recharging. Efficient energy management is becoming a priority for developers seeking to improve application performance and meet the needs of users in conditions of limited hardware resources.

Denis Shelmanov 

Within the framework of this work, methods of designing highly loaded systems based on a microservice architecture were studied, which allows the creation of flexible and scalable IT infrastructures to ensure resistance to intense loads. By splitting into independent modules, the microservice architecture improves reliability and reduces the dependency between services. The main objectives of the research include the development of a methodology for system decomposition, optimizing resource use, and ensuring fault tolerance, which is achieved through the use of Event Sourcing and CQRS patterns that provide state management and asynchronous event processing. Additionally, to achieve high adaptability and scalability, containerization, and orchestration of services through Kubernetes are used, which allows for dynamic load distribution and simplifies service management in conditions of intensive operation. The results of the analysis confirmed that the implementation of a microservice architecture minimizes system response time, improves data consistency, and reduces the risk of failures. In turn, the findings demonstrate that the microservice approach to the design of high-load systems provides flexibility, independent scaling, and adaptability, which is critically important for modern information platforms focused on continuous operation and prompt response to changes in load.

Gainetdionov Ainur Fanurovich 

This article discusses the use of pre-trained diffusion models for image segmentation. The study aims to increase the accuracy of segmentation and reduce required training data by using diffusion methods. During the work, models based on noise generation and removal were used, which allows for improvement in the quality of segmentation masks. The methodology includes the use of U-Net and Transformer algorithms, which contribute to the creation of high-precision segmentation masks of objects in images. The results showed that the use of pre-prepared models significantly improved the accuracy and consistency of segmentation masks compared to traditional methods. In conclusion, it is noted that the proposed approach provides higher segmentation efficiency without requiring extensive data, which opens up prospects for further technology development in various fields..

Raoul Hira

Eliminating the use of security questions and answers as an authentication or password recovery method has become an important step in improving the security of client accounts. Traditional security questions and answers based on user knowledge are criticized because of their vulnerability to guessing, social engineering attacks, or past security breaches. Such threats undermine their reliability as a means of protecting account access. Instead, modern approaches such as biometric authentication, multi-factor authentication (MFA), and password-free methods, including Magic Link, offer a higher level of security, reducing the risks of account compromise. The implementation of these methods requires a careful and step-by-step approach to minimize risks and increase the stability of security systems in the face of ever increasing threats and technological changes.

Avishkar Nikum

The article deals with the actual problem of countering phishing attacks in a corporate environment. The relevance of the topic is due to the rapid increase in the number and complexity of such attacks in the context of the digitalization of business processes, which leads to significant financial and reputational losses for business entities. The goal is to systematize modern scientific ideas in this field and develop an algorithm that allows you to protect your corporate information infrastructure from phishing threats. The study revealed contradictions between the need to strengthen security measures and maintain the operational efficiency of business processes. In addition, the discrepancies between the pace of development of protective technologies and the evolution of social engineering methods used by intruders are very significant. It was concluded that effective counteraction to phishing attacks requires integrating advanced technological solutions (machine learning, blockchain, quantum cryptography) with organizational measures, and carefully thought-out educational initiatives. Special attention is paid to the formation of a cybersecurity culture and the introduction of innovative authentication methods. The article is of interest to information security specialists, IT department heads, and cybersecurity researchers. The presented results can be used in the development and improvement of protection systems against phishing attacks in the corporate sector.

Inna Logunova

Integrating Artificial Intelligence (AI) into Business Intelligence (BI) is fundamentally transforming how organizations process data and make decisions. This article analyzes current trends in AI applications within BI, focusing on key innovations such as augmented analytics, generative AI assistants, automated machine learning models, and neural networks. These advancements enable non-technical users to extract insights independently, enhance data preparation, and facilitate informed decision-making, thereby democratizing access to data analytics. Despite these benefits, the adoption of AI-driven BI introduces significant challenges. These include accountability for AI-assisted decisions, ethical governance to prevent biases in AI models, cost implications associated with implementing large language models, and data security and privacy concerns. The article discusses the crucial role of Explainable AI (XAI) in providing transparency and building user trust. It also outlines strategies for mitigating biases, conducting cost benefit analyses, and establishing robust data governance policies. By addressing both the opportunities and the challenges, the article proposes comprehensive approaches for integrating AI-oriented solutions into organizational processes. This ensures consistent, ethical, and effective decision-making, aligning technological advancements with business objectives.

Asha Seshagiri

The article discusses aspects of multi-factor authentication (MFA) using biometric data, which is one of the most effective methods of protecting information systems from unauthorized access. Traditional authentication methods such as passwords are often vulnerable to various types of attacks, including phishing and brute force attacks. The introduction of biometric data, such as fingerprints, and face and voice recognition, significantly increases the level of security due to their uniqueness and complexity of forgery. The article highlights the advantages and challenges associated with the integration of biometrics into MFA systems, including technical and organizational aspects. Examples of successful application of biometric technologies in various fields are also considered and recommendations are given to improve their reliability and ease of use. Special attention is paid to adaptive authentication and password-free methods, which represent the future of cybersecurity.

Okhonko Pylyp

Protection methods against XSS attacks on the client side are key approaches to ensuring the security of web applications that prevent the introduction and execution of malicious code in users’ browsers. The main methods include data validation and filtering, using the Content Security Policy (CSP) protocol, encoding user input, as well as using mechanisms to detect and prevent the execution of suspicious scripts. These measures are aimed at reducing the risks of both reflected and persistent XSS attacks, providing a higher level of protection for confidential information. Special attention is paid to an integrated approach that includes the interaction of server and client protection methods, which minimizes vulnerabilities and increases the resistance of web applications to various types of attacks.

Kuzevanov Igor

In conditions of high dynamics of load on microservice architectures, managing the state of services becomes a critically important task. Traditional approaches do not always provide the necessary flexibility and adaptability, which requires the development of new management models and methods. This paper examines modern models designed to effectively manage the state of microservices, with an emphasis on their adaptation to changing load conditions. Various approaches to load balancing, automatic scaling, and fault tolerance are being evaluated. It is shown that the integration of machine learning methods and intelligent control systems can significantly increase the stability of the microservice architecture to load changes and minimize response time. As a result of the research, a comprehensive state management model for microservices is proposed, capable of adapting to dynamic conditions and ensuring high availability and system performance.