A model of ensuring LLM cybersecurity

Oleksii Neretin, Vyacheslav Kharchenko

Abstract


The subject of study is a model for ensuring cybersecurity of Large Language Models (LLM). The goal of this study is to develop and analyze the components of the LLM cybersecurity model to improve its assessment accuracy and ensure the required security level. Tasks: the abstract structure of LLM systems should be suggested and analyzed; a common model of cybersecurity of LLM systems (LLMS) should be built; a cybersecurity model of LLM as a main component of LLMS should be developed; the elements of the developed cybersecurity model should be analyzed; potential case studies should be described and an example of risk criticality analysis for one of the threats of the LLM should be provided; the directions of future research should be substantiated on the identification, classification, criticality analysis, and collection of exploits to test the stability of LLM. The research results: the basic high-level architecture of LLMS, which consists of external sources, the LLM service, server functions, and storage environments, is developed; a common LLM cybersecurity model was built based on this architecture; the cybersecurity model was developed, which is an independent component of the overall cybersecurity model of LLMS and is based on a chain of the following elements: threat, vulnerability, attack, risks, and countermeasures; in addition, an analysis of the elements of the LLM cybersecurity model is conducted, and a sequence of countermeasures is proposed. Conclusions. This study determines that improving the cybersecurity of LLM is an important and urgent task, given the widespread use of these models in many areas of human life. The importance of developing an LLM cybersecurity model is that it is the baseline for all subsequent research. The practical significance of analyzing the model’s elements lies in using them to conduct experiments to simulate cyber attacks on LLM. The main contributions of this study are the LLM and LLMS cybersecurity models, the formalization of the results of these experiments, an assessment of the criticality level for cyber risks of the models, and the choice of countermeasures based on the coefficient of their effectiveness. In this case, ensuring an acceptable risk level for LLM is possible at a minimal cost. Areas for further research: definition and classification of exploits to test LLM security; methodology for collecting these exploits; analysis of the criticality of the damage they cause for various applications.

Keywords


LLM; cybersecurity of LLM; cybersecurity model; threat; vulnerability; attack; risk; countermeasures

Full Text:

PDF

References


Clusmann, J., Kolbinger, F. R., Muti, H. S., Carrero, Z. I., Eckardt, J.-N., Laleh, N. G., Löffler, C. M. L., Schwarzkopf, S.-C., Unger, M., Veldhuizen, G. P., Wagner, S. J., & Kather, J. N. The future landscape of large language models in medicine. Communications medicine, 2023, vol. 3, no. 1, article no. 141. DOI: 10.1038/s43856-023-00370-1.

Nejjar, M., Zacharias, L., Stiehle, F., & Weber, I. LLMs for science: Usage for code generation and data analysis. Journal of Software: Evolution and Process, 2025, vol. 37, no. 1. DOI: 10.1002/smr.2723.

Wang, S., Xu, T., Li, H., Zhang, C., Liang, J., Tang, J., Yu, P. S., & Wen, Q. Large Language Models for Education: A Survey and Outlook. arXiv preprint arXiv:2403.18105, 2024. DOI: 10.48550/arXiv.2403.18105.

Li, Y., Zhao, H., Jiang, H., Pan, Y., Liu, Z., Wu, Z., Shu, P., Tian, J., Yang, T., Xu, S., Lyu, Y., Blenk, P., Pence, J., Rupram, J., Banu, E., Liu, N., Wang, L., Song, W., Zhai, X., Song, K., Zhu, D., Li, B., Wang, X., & Liu, T. Large Language Models for Manufacturing. arXiv preprint arXiv:2410.21418, 2024. DOI: 10.48550/arXiv.2410.21418.

Samma, H., & El-Ferik, S. UAV Visual Path Planning Using Large Language Models. Transportation Research Procedia, 2025, vol. 84, pp. 339-345. DOI: 10.1016/j.trpro.2025.03.081.

Hannig, L., Bush, A., Aksoy, M., Becker, S. & Ontrup, G. Campus AI vs Commercial AI: A Late-Breaking Study on How LLM As-A-Service Customizations Shape Trust and Usage Patterns. arXiv preprint arXiv:2505.10490, 2025. DOI: 10.48550/arXiv.2505.10490.

Kharchenko, V., Fesenko, H., & Illiashenko, O. Quality Models for Artificial Intelligence Systems: Characteristic-Based Approach, Development and Application. Sensors, 2022, vol. 22, no. 13, article no. 4865. DOI: 10.3390/s22134865.

Neretin, O., & Kharchenko, V. Zabezpechennya kiberbezpeky system shtuchnoho intelektu: analiz vrazlyvostey, atak i kontrzakhodiv [Ensurance of artificial intelligence systems cyber security: analysis of vulnerabilities, attacks and countermeasures]. Journal of Lviv Polytechnic National University. Information Systems and Networks, 2022, vol. 12, pp. 7-22. DOI: 10.23939/sisn2022.12.007. (In Ukrainian).

Wang, Y., Li, H., Han, X., Nakov, P., & Baldwin, T. Do-Not-Answer: A Dataset for Evaluating Safeguards in LLMs. arXiv preprint arXiv:2308.13387, 2023. DOI: 10.48550/arXiv.2308.13387.

Derner, E., Batistič, K., Zahálka, J., & Babuška, R. A Security Risk Taxonomy for Prompt-Based Interaction With Large Language Models. IEEE Access, 2024, vol. 12, pp. 126176-126187. DOI: 10.1109/ACCESS.2024.3450388.

Weidinger, L., Uesato, J., Rauh, M., Griffin, C., Huang, P.-S., Mellor, J., Glaese, A., Cheng, M., Balle, B., Kasirzadeh, A., Biles, C., Brown, S., Kenton, Z., Hawkins, W., Stepleton, T., Birhane, A., Hendricks, L. A., Rimell, L., Isaac, W., Haas, J., Legassick, S., Irving, G., & Gabriel, I. Taxonomy of Risks posed by Language Models. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency (FAccT '22), 2022, pp. 214–229. DOI: 10.1145/3531146.3533088.

Yao, Y., Duan, J., Xu, K., Cai, Y., Sun, Z., & Zhang, Y. A survey on large language model (LLM) security and privacy: The Good, The Bad, and The Ugly. High-Confidence Computing, 2024, vol. 4, iss. 2. DOI: 10.1016/j.hcc.2024.100211.

Cui, Т., Wang, Y., Fu, C., Xiao, Y., Li, S., Deng, X., Liu, Y., Zhang, Q., Qiu, Z., Li, P., Tan, Z., Xiong, J., Kong, X., Wen, Z., Xu, K., & Li, Q. Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems. arXiv preprint arXiv:2401.05778, 2024. DOI: 10.48550/arXiv.2401.05778.

Zhu, B., Mu, N., Jiao, J. & Wagner, D. Generative AI security: challenges and countermeasures. arXiv preprint arXiv:2402.12617, 2024. DOI: 10.48550/arXiv.2402.12617.

OWASP Top 10 for LLM Applications 2025. Available at: https://genai.owasp.org/llm-top-10/. (accessed 3.05.2025).

Babeshko, I., Illiashenko, O., Kharchenko, V., & Leontiev, K. Towards Trustworthy Safety Assessment by Providing Expert and Tool-Based XMECA Techniques. Mathematics, 2022, vol. 10, no. 13, p. 2297. DOI: 10.3390/math10132297.

Klondike, G. Threat Modeling LLM Applications. Available at: https://aivillage.org/large%20language%20models/2023/06/06/threat-modeling-llm. (accessed 3.05.2025).

Li, А., Zhou, Y., Raghuram, V. C., Goldstein, T., & Goldblum, M. Commercial LLM Agents Are Already Vulnerable to Simple Yet Dangerous Attacks. arXiv preprint arXiv:2502.08586, 2025. DOI: 10.48550/arXiv.2502.08586.

Wang, N., Walter, K., Gao, Y., & Abuadbba, A. Large Language Model Adversarial Landscape Through the Lens of Attack Objectives. arXiv preprint arXiv:2502.02960, 2025. DOI: 10.48550/arXiv.2502.02960.

Rehberger, J. Trust No AI: Prompt Injection Along The CIA Security Triad. arXiv preprint arXiv:2412.06090, 2024. DOI: 10.48550/arXiv.2412.06090.

Singh, A., & Gupta, B. B. Distributed Denial-of-Service (DDoS) Attacks and Defense Mechanisms in Various Web-Enabled Computing Platforms: Issues, Challenges, and Future Research Directions. International Journal on Semantic Web and Information Systems (IJSWIS), 2022, vol. 18, no. 1, pp. 1-43. DOI: 10.4018/IJSWIS.297143.

Lysenko, S., Bobrovnikova, K., Kharchenko, V., & Savenko, O. IoT Multi-Vector Cyberattack Detection Based on Machine Learning Algorithms: Traffic Features Analysis, Experiments, and Efficiency. Algorithms, 2022, vol. 15, no. 7, article no. 239. DOI: 10.3390/a15070239.

Hicks, M. T., Humphries, J., & Slater, J. ChatGPT is bullshit. Ethics and Information Technology, 2024, vol. 26, no. 2, pp. 1-10. DOI: 10.1007/s10676-024-09775-5.

Shah, C., Bender, E. M. Situating Search. In Proceedings of the 2022 Conference on Human Information Interaction and Retrieval, 2022, pp. 221-232. DOI: 10.1145/3498366.3505816.

Chollet, F. Deep Learning with Python, Second Edition. 2nd ed. Manning Publ., 2021. 504 p.

Das, B. C., Amini, M. H., & Wu, Y. Security and Privacy Challenges of Large Language Models: A Survey. ACM Computing Surveys, 2025, vol. 57, no. 6, pp. 1-39. DOI: 10.1145/3712001.

Du, H., Liu, S., & Cao, Y. Can Differentially Private Fine-tuning LLMs Protect Against Privacy Attacks? arXiv preprint arXiv:2504.21036, 2025. DOI: 10.48550/arXiv.2504.21036.

Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C. L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P., Leike, J., & Lowe, R. Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155, 2022. DOI: 10.48550/arXiv.2203.02155.

Mao, Y., He, J., & Chen, C. From Prompts to Templates: A Systematic Prompt Template Analysis for Real-world LLMapps. arXiv preprint arXiv:2504.02052, 2025. DOI: 10.48550/arXiv.2504.02052.

Perez, F., & Ribeiro, I. Ignore Previous Prompt: Attack Techniques For Language Models. arXiv preprint arXiv:2211.09527, 2022. DOI: 10.48550/arXiv.2211.09527.

Greshake, K., Abdelnabi, S., Mishra, S., Endres, C., Holz, T., & Fritz, M. Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection. In Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security, 2023, pp. 79-90. DOI: 10.1145/3605764.3623985.

Wei, A., Haghtalab, N., & Steinhardt, J. Jailbroken: How Does LLM Safety Training Fail? arXiv preprint arXiv:2307.02483, 2023. DOI: 10.48550/arXiv.2307.02483.

Siska, C., & Sankaran, A. AttentionDefense: Leveraging System Prompt Attention for Explainable Defense Against Novel Jailbreaks. arXiv preprint arXiv:2504.12321, 2025. DOI: 10.48550/arXiv.2504.12321.

Lin, L., Mu, H., Zhai, Z., Wang, M., Wang, Y., Wang, R., Gao, J., Zhang, Y., Che, W., Baldwin, T., Han, X., & Li, H. Against The Achilles' Heel: A Survey on Red Teaming for Generative Models. Journal of Artificial Intelligence Research, 2025, vol. 82, pp. 687-775. DOI: 10.1613/jair.1.17654.

Illiashenko, O., Kharchenko, V., Babeshko, I., Fesenko, H., & Di Giandomenico, F. Security-Informed Safety Analysis of Autonomous Transport Systems Considering AI-Powered Cyberattacks and Protection. Entropy, 2023, vol. 25, no. 8, article no. 1123. DOI: 10.3390/e25081123.

Vulnerability Metrics. Available at: https://nvd.nist.gov/vuln-metrics/cvss. (accessed 3.05.2025).

Bitton, R., Maman, N., Singh, I., Momiyama, S., Elovici, Y., & Shabtai, A. Evaluating the Cybersecurity Risk of Real-world, Machine Learning Production Systems. ACM Computing Surveys, 2023, vol. 55, no. 9, pp. 1-36. DOI: 10.1145/3559104.

Zemlianko, H., & Kharchenko, V. IMECA-analiz kiberbezpeky system bahatofunktsionalnykh flotiv BPLA pry kombinovanykh atakakh: bazovi modeli ta vybir kontrzakhodiv [IMECA analysis of cybersecurity for multi-functional UAV fleets under combined attacks: basic models and countermeasure choice]. Measuring and computing devices in technological processes, 2023, no. 4, pp. 225-233. DOI: 10.31891/2219-9365-2023-76-30. (In Ukrainian).

Zemlianko, H., & Kharchenko, V. Cybersecurity risk analysis of multifunctional UAV fleet systems: a conceptual model and IMECA-based technique. Radioelectronic and Computer Systems, 2023, vol. 0, no. 4, pp. 152-170. DOI: 10.32620/reks.2023.4.11.

Syed, N., Anwar, A., Baig, Z., & Zeadally, S. Artificial Intelligence as a Service (AIaaS) for Cloud, Fog and the Edge: State-of-the-Art Practices. ACM Computing Surveys, 2025, vol. 57, no. 8. DOI: 10.1145/3712016.

Javaid, S., Fahim, H., He, B. & Saeed, N. Large language models for UAVs: Current state and pathways to the future. IEEE Open Journal of Vehicular Technology, 2024, vol. 5, pp. 1166-1192. DOI: 10.1109/OJVT.2024.3446799.

Chao, P., Debenedetti, E., Robey, A., Andriushchenko, M., Croce, F., Sehwag, V., Dobriban, E., Flammarion, N., Pappas, G. J., Tramer, F. & Hassani, H. Jailbreakbench: An open robustness benchmark for jailbreaking large language models. arXiv preprint arXiv:2404.01318, 2024. DOI: 10.48550/arXiv.2404.01318.

Laktionov, O., Shefer, O., Laktionova, I., Halai, V., & Podorozhniak, A. Implementation of unsupervised learning models for analyzing the state's security level. Advanced Information Systems, 2024, vol. 8, no. 3, pp. 85-91. DOI: 10.20998/2522-9052.2024.3.10.

Rajakumareswaran, V., Raguvaran, S., Chandrasekar, V., Rajkumar, S., & Arun, V. DeepFake detection using transfer learning-based Xception model. Advanced Information Systems, 2024, vol. 8, no. 2, pp. 89-98. DOI: 10.20998/2522-9052.2024.2.10.

Yang, Z., Zhang, Y., Zeng, J., Yang, Y., Jia, Y., Song, H., Lv, T., Sun, Q., & An, J. AI-Driven Safety and Security for UAVs: From Machine Learning to Large Language Models. Drones, 2025, vol. 9, no. 6, article no. 392. DOI: 10.3390/drones9060392.

Sezgin, A. Scenario-Driven Evaluation of Autonomous Agents: Integrating Large Language Model for UAV Mission Reliability. Drones, 2025, vol. 9, no. 3, article no. 213. DOI: 10.3390/drones9030213.




DOI: https://doi.org/10.32620/reks.2025.2.13

Refbacks

  • There are currently no refbacks.