9004 2071-8217 Спецвыпуск 2025 1-189

RAR RUS 10-22 0000-0002-7485-4848 Peter the Great St. Petersburg Polytechnic University Aleksandrova Elena aleksandrova_eb@spbstu.ru 0009-0002-4454-327X Peter the Great St. Petersburg Polytechnic University Prokofieva Anna anna.prokofieva01@mail.ru Integration of guidance-based recovery mechanism into the learning with errors-based digital signature The article considers modern approaches to modification of digital signature schemes on lattices. Learning with errors (LWE) problem and its modifications using hints and rounding mechanisms to reduce computational costs and decrease the signature size without compromising security are estimated. An adapted scheme using the hint mechanism and optimized rounding based on the GLYPH signature protocol is proposed. A prototype is implemented and tested, the results of which demonstrate a decrease in the average time of signature generation and verification by ~1.4 times, and a decrease in the signature size by 14 % compared to the baseline scheme. 10.48612/jisp/2hfa-tvup-b6v9 004.056 Post-quantum cryptography Learning with Errors (LWE) hint rounding rejection sampling https://jisp.spbstu.ru/article/2025.22.1/ 2025_spetsvipusk-5-6.pdf

RAR RUS 23-33 0000-0002-7485-4848 Peter the Great St. Petersburg Polytechnic University Aleksandrova Elena aleksandrova_eb@spbstu.ru 0009-0002-0836-5920 Peter the Great St. Petersburg Polytechnic University Samareva Daria samarevadaria@gmail.com Lattice-based commitement scheme for proving linear relations over hidden values A hybrid lattice-based commitment scheme for anonymous proof of linear relations between hidden values is proposed. The proposed approach is based on a modification of the BDLOP zero-knowledge scheme, where Learning with Errors problem was replaced with Learning with Rounding problem, which reduced the parameter sizes and complexity of parameter sampling process. The proposed scheme retains its mathematical properties, including additive homomorphism, enabling it use for proving linear relations. The results obtained can be applied to the design of protocol of electronic voting and anonymous transactions. 10.48612/jisp/f7p4-n9p1-gtt6 004.056 Post-quantum cryptography lattices commitment scheme zeroknowledge proof learning with rounding learning with error https://jisp.spbstu.ru/article/2025.22.2/ 2025_spetsvipusk-5-6.pdf

RAR RUS 34-47 0009-0006-7175-8427 Peter the Great St. Petersburg Polytechnic University Bolokan Amvrosii bolokan.av@mail.ru 0000-0001-9659-1244 Peter the Great St. Petersburg Polytechnic University Poltavtseva Maria potavtseva@ibks.spbstu.ru Countering illegitimate smart voice assistant triggering This paper is about an ensuring the security of smart voice assistants against the most significant threats by reducing the number of false triggers. The paper analyses the threats to smart voice assistants and presents a list of their features. The goal is to reduce the number of unnecessary activations of smart voice assistant. An architecture and method for a custom security module that reduces false alarms are described. This security module was developed and tested with positive results. 10.48612/jisp/pn2v-vnu9-z3v4 004.056 Smart voice assistant voice biometrics dolphin attack information security threat model impersonalization white noise speech synthesis https://jisp.spbstu.ru/article/2025.22.3/ 2025_spetsvipusk-5-6.pdf

RAR RUS 48-57 0009-0008-8810-6307 Peter the Great St. Petersburg Polytechnic University Bondarenko Timur bondarenko.ta@edu.spbstu.ru 0000-0002-2009-5460 Peter the Great St. Petersburg Polytechnic University Ovasapyan Tigran otd@ibks.spbstu.ru 0009-0009-5163-9975 Peter the Great St. Petersburg Polytechnic University Piskov Aleksandr brontosd2@gmail.com Using entropy metrics to detect data integrity attacks in real-time Existing methods of detecting attacks on data integrity on file systems are investigated. A method of detecting such attacks based on the use of several entropy metrics is proposed. The efficiency of the proposed method is evaluated on the example of detection of existing ransomware. 10.48612/jisp/78up-h9mu-rmxt 004.056 Data integrity entropy dynamic analysis ransomware encryptor driver filter https://jisp.spbstu.ru/article/2025.22.4/ 2025_spetsvipusk-5-6.pdf

RAR RUS 58-68 0000-0001-9862-1507 Peter the Great St. Petersburg Polytechnic University Dakhnovich Andrey add@ibks.spbstu.ru 0009-0000-3249-4103 Peter the Great St. Petersburg Polytechnic University Bogina Vasilisa bogina_vm@spbstu.ru 0009-0008-9034-1365 Peter the Great St. Petersburg Polytechnic University Makeeva Anna makeeva.aa@edu.spbstu.ru Application of large language models in event forecasting field This article presents a study on the use of large language models (LLMs) for event prediction through the application of LLM agents - autonomous systems that utilize LLMs for reasoning, decision-making, and interaction with the environment. Various architectures of LLM agents are analyzed, including cooperative systems (ChatDev, MetaGPT), multi-agent debates (MAD, ChatEval), agents for web-based tasks (WebAgent, WebVoyager), and simulation-based agents (Generative Agents, EconAgent). Special attention is given to the features of predictive modeling powered by LLMs, where traditional approaches (regression, time series) are replaced by agent-based modeling and prompt engineering. The article presents experimental results on forecasting the outcome of a selected conflict using LLM agents (Mistral, DeepSeek) and the Retrieval-Augmented Generation (RAG) approach, based on data from analytical agencies, opinion leaders, and news sources. The study identifies a convergence of predictive assessments across polarized sources and outlines key requirements for forecasting systems: weighting sources by expert relevance, filtering out neutral data, and balancing the dataset. Additionally, the article formulates criteria for selecting data to be evaluated by simulation-based LLM agents. 10.48612/jisp/xmg2-m34p-rmn3 004.056 Artificial intelligence generative models large language models agentbased modeling social simulations LLM-agents natural language processing RAG prompt-engeneering conflict forecasting https://jisp.spbstu.ru/article/2025.22.5/ 2025_spetsvipusk-5-6.pdf

RAR RUS 69-81 0009-0009-0312-6243 Peter the Great St. Petersburg Polytechnic University Lavrenko Egor egor.lavr1@gmail.com 0000-0001-9659-1244 Peter the Great St. Petersburg Polytechnic University Poltavtseva Maria potavtseva@ibks.spbstu.ru Data processing and mining to detect a data privacy violations from internal intruder in a DBMS The paper is devoted to the detection of intrusions and violations regarding the confidentiality of data stored in a database based on behavioral analysis. A particular difficulty in this area is taking into account not only the query syntax, but also the semantic relationships of thedata, since syntactic and contextual approaches do not allow detecting all types of attacks. Based on the analysis of well-known studies, a method is proposed for detecting anomalies in user behavior based on author’s metrics for evaluating behavior and the coverage of requested data. The proposed method develops the well-known research, but at the same time significantly surpasses it in the task of detecting certain types of behavioral abnormalities. An important part of the work is to identify the application features of this type of analysis and its limitations. 10.48612/jisp/gr29-tnf6-db9d 004.056 Data management systems data security data mining behavior analysis data privacy data processing anomaly detection insider detection https://jisp.spbstu.ru/article/2025.22.6/ 2025_spetsvipusk-5-6.pdf

RAR RUS 82-98 0000-0003-2849-4682 Peter the Great St. Petersburg Polytechnic University Lavrova Daria lavrova_ds@spbstu.ru

Russia, 195251, St. Petersburg, Polytechnicheskaya str., 29

Peter the Great St. Petersburg Polytechnic University Zyryanova Anastasiya zyryanova.aa@edu.spbstu.ru 0009-0008-4593-4444 Peter the Great St. Petersburg Polytechnic University Lyrchikov Alexander lyrchikov.aa@gmail.com

Russia, 195251, St. Petersburg, Polytechnicheskaya str., 29

Detection of fake news spread using machine learning The article addresses the problem of detecting unreliable news content and proposes a solution based on machine learning methods. Modern approaches to assessing the credibility of textual and multimedia content have been analyzed, promising techniques have been identified and adapted for the Russian-language media environment. A combined method for detecting fake news is proposed, based on the joint analysis of textual and multimedia data, as well as content distribution patterns. Testing of the proposed method has confirmed its effectiveness and applicability for automated detection of unreliable news content in real-world information systems. 10.48612/jisp/7eua-z1tp-4pzk 004.056 Fake news detection misinformation spreading machine learning text analysis news content filtering news classification natural language processing https://jisp.spbstu.ru/article/2025.22.7/ 2025_spetsvipusk-5-6.pdf

RAR RUS 99-111 0009-0008-4593-4444 Peter the Great St. Petersburg Polytechnic University Lyrchikov Alexander lyrchikov.aa@gmail.com

Russia, 195251, St. Petersburg, Polytechnicheskaya str., 29

0000-0001-9862-1507 Peter the Great St. Petersburg Polytechnic University Dakhnovich Andrey add@ibks.spbstu.ru Peter the Great St. Petersburg Polytechnic University Moskvin Dmitry moskvin_da@spbstu.ru An automated method for assessing the correctness of distributed algorithms using chaos engineering testing In the paper a method for assessing the correctness of distributed algorithms using chaos engineering techniques to enhance their testing efficiency is presented. An analysis of current research in the field of testing distributed systems - such as federated learning systems for Large Language Models (LLMs) - and chaos engineering is provided. Existing chaos engineering-based testing methods and tools are analyzed, and their shortcomings are identified. As a result, a method for assessing the correctness of distributed systems through chaos engineering testing has been developed; vulnerability testing in open-source projects was conducted, including a comparison with existing approaches. The obtained results confirm the effectiveness of the proposed method. 10.48612/jisp/ed6x-521z-23h7 004.056 Testing chaos engineering fuzzing LLM federated learning distributed algorithms distributed systems https://jisp.spbstu.ru/article/2025.22.8/ 2025_spetsvipusk-5-6.pdf

RAR RUS 112-123 0000-0003-1345-1874 Peter the Great St. Petersburg Polytechnic University Pavlenko Evgeny pavlenko_eyu@spbstu.ru 0009-0002-7009-2265 Peter the Great St. Petersburg Polytechnic University Anoshkin Ilya ilya.anoschkin@outlook.com Automated security analysis of software for the Android operating system This paper presents methods of automated security analysis of Android applications, which can be used to search for cryptographic vulnerabilities, vulnerabilities of third-party software components, authentication, authorization, as well as to detect the storage and transmission of sensitive information in plaintext. The accuracy of search for the given types of vulnerabilities by means of automated vulnerability search and software prototype is analyzed. 10.48612/jisp/6h6a-6kzh-84g3 004.056 Android vulnerability scanning dynamic instrumentation security standard https://jisp.spbstu.ru/article/2025.22.9/ 2025_spetsvipusk-5-6.pdf

RAR RUS 124-133 0009-0006-6856-2108 Peter the Great St. Petersburg Polytechnic University Pahomov Maksim pahomov_ma@spbstu.ru The architecture of a software package for wireless ad-hoc network protection from active network attacks Software solutions for active network attack detection and prevention on wireless ad-hoc networks have been analyzed. The requirements for the architecture of a software system designed to protect wireless ad-hoc networks from active network attacks have been formulated. The architecture meeting the requirements is proposed. A software prototype has been developed that implements the proposed solution, and its evaluation has been performed. 10.48612/jisp/frpt-96b9-rn4t 004.056 Information security ad-hoc networks network attack prevention intrusion detection systems https://jisp.spbstu.ru/article/2025.22.10/ 2025_spetsvipusk-5-6.pdf

RAR RUS 134-144 0009-0008-2356-7487 Peter the Great St. Petersburg Polytechnic University Prisich Maria prisichma@gmail.com 0000-0003-0374-4649 Peter the Great St. Petersburg Polytechnic University Bezborodko Anastasia nsspbpoly@gmail.com The use of convolutional neural networks to enhance the security of steganographic methods The paper proposes the approach to increase the secrecy of steganography methods in images using convolutional neural networks (CNN). CNNs are integrated into the data embedding process and allow you to minimize the traces of concealment that can be detected by stegoanalyzers. Two implementation options are considered: based on the least significant bit (LSB) and discrete cosine transform (DCT) methods, and also their modifications using CNN. The task of ensuring the secrecy and reliability of embedding was solved in stages: the visual quality of the images, the reliability of message extraction and the resistance to detection by classical methods of stegoanalysis were analyzed. The results of quality and secrecy assessment experiments have confirmed the effectiveness of the proposed approach. 10.48612/jisp/du6k-pfkx-7ngv 004.056 Steganography LSB DCT machine learning convolution neural networks stegoanalysis https://jisp.spbstu.ru/article/2025.22.11/ 2025_spetsvipusk-5-6.pdf

RAR RUS 145-155 0009-0000-5788-8619 Peter the Great St. Petersburg Polytechnic University Rochman Nikita nik.roxman@mail.ru 0000-0003-0374-4649 Peter the Great St. Petersburg Polytechnic University Bezborodko Anastasia nsspbpoly@gmail.com 0000-0002-7485-4848 Peter the Great St. Petersburg Polytechnic University Aleksandrova Elena aleksandrova_eb@spbstu.ru Hiding data in audio files using coverless steganography methods The paper proposes a coverless steganographic approach for hiding data in audio files. It is based on an architecture that includes the generative neural network model RealNVP for processing streaming data and auxiliary encoding and decoding modules. Piano-jazz was chosen as a musical genre for generating audio files. The results of experimental studies have confirmed the effectiveness of using streaming neural networks for the tasks of coverless steganography in audio files. 10.48612/jisp/1ng9-14k5-a587 004.056 Steganography coverless steganography generative model RealNVP model audio file generation https://jisp.spbstu.ru/article/2025.22.12/ 2025_spetsvipusk-5-6.pdf

RAR RUS 156-167 0009-0007-2389-0789 St. Petersburg State Marine Technical University Fokina Sofia sofiya.fockina@gmail.com St. Petersburg State Marine Technical University Yakovleva Polina iakovleva.polina16@mail.ru 0000-0001-6695-2328 St. Petersburg State Marine Technical University Garkushev Alexander sangark@mail.ru 0009-0001-8806-787X St. Petersburg State Marine Technical University Morozova Anna amorozova94@gmail.com 0000-0001-9665-0128 Peter the Great St. Petersburg Polytechnic University Suprun Alexander afs54@inbox.ru

Russia, 195251, St. Petersburg, Polytechnicheskaya str., 29

Information model for countering the illegal distribution of personal data in information systems The article presents an information model for ensuring the security of personal data in information systems, based on the Secure Remote Password protocol and Russian cryptographic algorithms GOST R34.12-2015 (“Kuznechik”) and GOST 34.11-2018 (“Stribog”). An analysis of threats and vulnerabilities of the information systems, the regulatory framework and modern methods of protecting personal data is carried out. A modular software implementation has been developed that is resistant to the main types of attacks, including traffic interception, man-in-the-middle attacks and database leaks. 10.48612/jisp/1pk8-r79m-xm3m 004.65 Personal data information security SRP protocol cryptographic algorithms MITM-attacks data protection https://jisp.spbstu.ru/article/2025.22.13/ 2025_spetsvipusk-5-6.pdf

RAR RUS 168-178 Peter the Great St. Petersburg Polytechnic University Tsibulskas Konstantin tsibulskas_ka@spbstu.ru 0000-0002-2264-7513 Peter the Great St. Petersburg Polytechnic University Krundyshev Vasiliy krundyshev_vm@spbstu.ru 0000-0002-9732-0099 Peter the Great St. Petersburg Polytechnic University Kalinin Maxim max@ibks.spbstu.ru

Russia, 195251, St. Petersburg, Polytechnicheskaya str., 29

Ensuring the stability of online learning artificial intelligence systems based on model similarity assessment The paper studies the problem of protecting artificial intelligence systems with online learning from poisoning attacks. To improve the stability, an approach is proposed based on assessing the similarity of the operation of two computational models: the reference (initial) and the operational (test). The following indicators of stability violation were identified: a decrease in the total accuracy (TA), total prediction value (TPV), and a decrease in the cosine similarity of model weights (cos_similarity). As a result of experimental study, it was found that the proposed solution allows for timely detection of poisoned data, maintaining high classification accuracy during targeted attacks on the computational model, which is further trained on test data. 10.48612/jisp/ad7f-mgh1-urdh 004.056 Poisoning attack artificial intelligence security online learning model similarity assessment https://jisp.spbstu.ru/article/2025.22.14/ 2025_spetsvipusk-5-6.pdf

RAR RUS 179-188 0009-0009-7468-2511 Peter the Great St. Petersburg Polytechnic University Yudin Sergey yudin.sa@spbstu.ru 0000-0002-2009-5460 Peter the Great St. Petersburg Polytechnic University Semenov Pavel semenov_po@ibks.spbstu.ru 0009-0008-4442-5365 Peter the Great St. Petersburg Polytechnic University Kubrin Georgiy kubrin@ibks.spbstu.ru Automating of container images security scanning and analysis The structure and composition of container images, as well as the related security issues, are analyzed. The existing scanning methods for detecting vulnerabilities in container images are analyzed, their advantages and disadvantages are highlighted. An approach addressing the identified shortcomings is proposed. A software prototype of an automated security scanning system for container images with support for dynamic monitoring is developed and tested. 10.48612/jisp/kg63-a9kb-r12f 004.056 Containers container images container image vulnerability scanners Docker https://jisp.spbstu.ru/article/2025.22.15/ 2025_spetsvipusk-5-6.pdf