9004 2071-8217 Спецвыпуск 2026 1-105

RAR RUS 8-19 0009-0004-0955-1527 Saint Petersburg Electrotechnical University Amenitsky Alexey arbat365@mail.ru Proactive AI risk management The article examines the current problem of cryptocurrency fraud escalation, characterized by the abuse of advanced technological terms (“quantum artificial intelligence” and “cyber immunity”) to legitimize  phishing platforms and social engineering schemes. Quantum AI-SecOps Cyber Immunity Framework (QAS-CIF), which combines the principles of bio-inspired cyber immunity, continuous telemetry Security Operations (SecOps), and quantum machine learning (QML) algorithms. The paper formalizes the four architectural pillars of the framework, develops a matrix for neutralizing specific attack vectors of cryptocurrency fraudsters, and proposes a three-level model of maturity for the implementation of this architecture. The results of the study demonstrate that the transition from reactive protection to a proactive system capable of autonomously isolating “pathogenic” digital objects is critically necessary for the protection of financial assets in the quantum era. 004.056 Cyber immunity quantum artificial intelligence Security Operations (SecOps) cryptocurrency fraud deepfake detection zero trust architecture quantum machine learning SOAR https://jisp.spbstu.ru/article/2026.27.1/ s_2026.png

RAR RUS 20-28 0009-0009-1545-8835 St. Petersburg State Marine Technical University Vyvolokina Albina albina.vyvolokina@mail.ru Methodology for identifying the authorship of objects generated by artificial intelligence The article is devoted to the actual problem of identifying the authorship of objects created using artificial intelligence technologies. In the context of the rapid rise of deepfakes and other synthetic content, along with the proliferation of anonymous or fake accounts on artificial intelligence platforms, there is an urgent need for reliable and legally significant methods to determine who is responsible for generating a given object. The author presents a method for AI-generated object authorship identification in order to increase the transparency and security of the circulation of such objects, which is based on fingerprint and SynthID technologies. 347.78:004.056:004.8 Deepfakes artificial intelligence cybersecurity unique identifier identification authorship https://jisp.spbstu.ru/article/2026.27.2/ s_2026.png

RAR RUS 30-40 7006566675 0000-0002-6076-7241 Emperor Alexander I St. Petersburg State Transport University Kornienko Anatoliy kaa.pgups@yandex.ru

Russia, 190031, St. Petersburg, Moskovsky ave., 9

0000-0003-2683-0697 Emperor Alexander I St. Petersburg State Transport University Kornienko Svetlana sv.diass99@yandex.ru St. Petersburg State Transport University of Emperor Alexander I Gzhibovskiy Ivan gzhibovskiy02@mail.ru Developing a spam filter using artificial intelligence The paper examines existing methods of protecting corporate e-mail from spam and analyzes approaches to its detection. Practice-oriented requirements for spam filtering systems are formulated, as well as a comparative study of machine learning models within a single reproducible experimental circuit. A software solution for spam filtering based on machine learning algorithms has been developed and tested. It has been established that, given the selected text data processing scheme-specifically the use of TF-IDF vectorization and class balancing-the SVM algorithm demonstrates the optimal balance between classification quality and operational efficiency. The obtained results can serve as a foundation for the further development of hybrid filtering systems. 004.056 Spam machine learning information security https://jisp.spbstu.ru/article/2026.27.3/ s_2026.png

RAR RUS 41-50 0009-0000-1660-7809 State Marine Technical University Lopatin Maxim mslopatin01@mail.ru 0000-0001-6695-2328 St. Petersburg State Marine Technical University Garkushev Alexander sangark@mail.ru 0000-0001-9665-0128 Peter the Great St. Petersburg Polytechnic University Suprun Alexander afs54@inbox.ru

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

Application of distribution laws to model the influence of the human factor The article resolves the contradiction between the need for proactive quantitative information security risk management and the static nature of traditional qualitative methods that ignore the stochastic nature of human behavior. A method of differentiated probabilistic threat modeling using Poisson, Gauss, and uniform law distributions is proposed to evaluate unintended errors, intentional actions, and violations of cyber hygiene, respectively. The results of the computational experiment confirm the statistically significant superiority of the proposed approach over classical matrix methods. The work creates the basis for the transition to a quantified allocation of protective resources and the integration of models into real-time monitoring systems. 004.056.5 Human factors information security risk modeling Poisson distribution normal distribution risk assessment https://jisp.spbstu.ru/article/2026.27.4/ s_2026.png

RAR RUS 51-60 Peter the Great St. Petersburg Polytechnic University Moskalev Nikita moskalev.no@ibks.spbstu.ru Peter the Great St. Petersburg Polytechnic University Logacheva Svetlana logacheva_sv@edu.spbstu.ru 0000-0003-2849-4682 Peter the Great St. Petersburg Polytechnic University Lavrova Daria lavrova_ds@spbstu.ru

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

Protection of artificial intelligence models from backdoor attacks The paper is devoted to the problem of protecting artificial intelligence models from backdoor attacks. During the work, the methods of protecting artificial intelligence models from this type of attack were analyzed. The identified shortcomings of these methods confirm the lack of a solution that preserves model quality while providing an enough level of protection against this class of attacks. To address this issue, an approach was developed to protect artificial intelligence models from backdoor attacks. Its feasibility was theoretically substantiated and empirically confirmed. The results obtained can be used for further research and the development of new security solutions. 004.056 Artificial intelligence neural networks classification poisoning if AI models backdoor attack https://jisp.spbstu.ru/article/2026.27.5/ s_2026.png

RAR RUS 61-69 0000-0002-1143-5275 V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences Salomatin Aleksandr sandr@ipu.ru 0000-0003-1316-3346 V. A. Trapeznikov Institute of Control Sciences of Russian Academy of Sciences Davydov Vyacheslav davydov@ipu.ru A model for analyzing heterogeneous behavioral digital fingerprints based on structurally sound distance metrics The paper proposes a model for analyzing heterogeneous behavioral digital fingerprints based on a structurally sound distance metrics selection. Adaptive weighting ensures a correct comparison of binary, quantitative, cyclic features, and time series within a unified space. Experimental validation of the model on the public Behaviour Biometrics Dataset demonstrated that a sound choice of the threshold value ensures a rational balance between quality evaluation metrics: the true acceptance rate of legitimate sessions E1 = 1.000, the rate of detected anomaly illegitimate sessions E2 = 0.0057, with the failure-to-recognize rate E3 = 0.0028. The model is designed not for autonomous decision-making, but for transmitting a structured risk signal to an external system control loop, where the threshold value can be adaptively adjusted to meet specific objectives. 004.056.5 Behavioral digital footprint adaptive metric selection feature taxonomy classification with rejection Fisher’s index https://jisp.spbstu.ru/article/2026.27.6/ s_2026.png

RAR RUS 74-80 57200960264 0000-0001-6289-3295 Russian State Hydrometeorological University Sikarev Igor sikarev@yandex.ru

Russia, 192007, St. Petersburg, Voronezhskaya str., 79

0000-0003-0554-5790 Admiral Makarov State University of Maritime and Inland Shipping Abramov Valery val.abramov@mail.ru Russian State Hydrometeorological University Bolshakov Vladimir v.a.bolsh@mail.ru Russian State Hydrometeorological University Vekshina Tatiana t.v.vekshina@mail.ru Russian State Hydrometeorological University Korinets Ekaterina miffi89@mail.ru Automation of processes geoinformation water management with mind information security There is proposed conceptual formulation for directions of automation of processes geoinformation management within water management, including issues of collection, processing and visualization of hydrometeorological and hydrological data. There are used methods of geoinformation management, system analysis, open source scanning technologies, technologies for automated processing of big data, technologies for developing web tools, methods and technologies for ensuring information security. All data is taken from open sources. When automating the processes of geoinformation management of water management facilities, including shipping, it is proposed to use geoinformation online platforms. It has been established that when planning automation activities, information security requirements must also be taken into account 003.26 Automation geoinformation management water management shipping information security https://jisp.spbstu.ru/article/2026.27.7/ s_2026.png

RAR RUS 81-90 0009-0008-8784-8720 Peter the Great St. Petersburg Polytechnic University Sobolev Nikolay sobolev.nv@edu.spbstu.ru Systematization of zerodynamic attacks for different types of cyber-physical systems Zero-dynamics attacks are a class of covert attacks on cyber-physical systems in which a change in the internal state of an object is not accompanied by a comparable deviation in the monitored output signals. This complicates their detection using traditional monitoring tools based on the analysis of measurements and residual signals. The aim of this paper is to systematize the main types of zero-dynamics attacks and identify the specifics of their implementation in various classes of cyber-physical systems. Five types of zero-dynamics attacks are considered and then compared. Furthermore, industrial, energy, transport, and aerospace cyber-physical systems are identified, for which the applicability of these attack types is analyzed 004.056 Cyber-physical system zero-dynamic attack sampling zeros nonlinear system https://jisp.spbstu.ru/article/2026.27.8/ s_2026.png

RAR RUS 96-104 0009-0006-5861-2511 Peter the Great St. Petersburg Polytechnic University Sorokin Alexandr mkentrru@yandex.ru 0000-0003-1345-1874 Peter the Great St. Petersburg Polytechnic University Pavlenko Evgeny pavlenko_eyu@spbstu.ru Application of machine learning to the problems of recovering the sequence of deleted data fragments This paper examines a significant problem in digital forensics: recovering fragmented files. A graph representation of the NTFS file system, reflecting the characteristics of its fragmentation process, is created. A method for defragmenting deleted NTFS data by analyzing the file system state using machine learning is developed. This method is based on embedding the graph representation using a relational graph neural network. An experimental evaluation of the developed method is conducted in comparison with the PhotoRec file carving tool using open data 004.056 Data recovery deleted files defragmentation graph neural networks NTFS filesystem graph representation of the file system https://jisp.spbstu.ru/article/2026.27.9/ s_2026.png