Journals / NATURENGS / 2026 / Volume: 6 Issue: 1

Depth Analysis of Vulnerabilities in Training and Inference Times of Large Language Models

Canan Batur Şahin(MALATYA TURGUT ÖZAL ÜNİVERSİTESİ)0000-0002-2131-6368
Journal
NATURENGS
Pages
—
Publication Date
2025-06-29
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

The proposed defensive mechanisms showcase impressive model performance, highlighted by consistent successful evaluation outcomes.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Large language models (LLMs) have dramatically reshaped the field of natural language processing, presenting groundbreaking advancements in many areas, from chatbots to content creation.
📊
Key Findings & Data:Our experimental results across multiple attack vectors demonstrate varying degrees of model susceptibility and reveal critical security implications.
💡
Conclusion & Impact:The proposed defensive mechanisms showcase impressive model performance, highlighted by consistent successful evaluation outcomes.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Large Language Models (LLMs) · Cyber-security · Adversarial attacks

Abstract

Large language models (LLMs) have dramatically reshaped the field of natural language processing, presenting groundbreaking advancements in many areas, from chatbots to content creation. However, with the increasing adoption of these sophisticated models, it is crucial to scrutinize the vulnerabilities associated with their training and inference stages. This comprehensive analysis highlights the critical threats and inefficiencies inherent to these processes and emphasizes the need for robust countermeasures. This paper presents an extensive study of training and inference time vulnerabilities in Large Language Models (LLMs), specifically focusing on poisoning, backdoor, paraphrasing, and spoofing attacks. We introduce novel evaluation frameworks and detection mechanisms for each attack type. Our experimental results across multiple attack vectors demonstrate varying degrees of model susceptibility and reveal critical security implications. The proposed defensive mechanisms showcase impressive model performance, highlighted by consistent successful evaluation outcomes.

🔖

Cite This Article

Şahin, C. B. (2025). Depth Analysis of Vulnerabilities in Training and Inference Times of Large Language Models. NATURENGS, Volume: 6 Issue: 1. https://doi.org/10.46572/naturengs.1636277

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (16)

  1. Zhang, E. Y., Cheok, A. D., Pan, Z., Cai, J., & Yan, Y. (2023). From Turing to Transformers: A Comprehensive Review and Tutorial on the Evolution and Applications of Generative Transformer Models. Sci, 5(4), 46. https://doi.org/10.3390/sci5040046
  2. Yang,J. (2024). Large language models privacy and security. Applied and Computational Engineering,76,177-188.
  3. Guven, M. (2024). A Comprehensive Review of Large Language Models in Cyber Security. International Journal of Computational and Experimental Science and Engineering, 10(3). https://doi.org/10.22399/ijcesen.469
  4. OWASP. (2024). OWASP Top 10 for Large Language Model Applications. OWASP Foundation.
  5. B. S. Latibari et al., (2024). Transformers: A Security Perspective, IEEE Access, vol. 12, pp. 181071-181105, doi: 10.1109/ACCESS.2024.3509372.
  6. Du, W., Li, P., Li, B., Zhao, H., & Liu, G. (2023). UOR: Universal Backdoor Attacks on Pre-trained Language Models. ArXiv. https://arxiv.org/abs/2305.09574.
  7. Zhang, Y., Rando, J., Evtimov, I., Chi, J., Smith, E. M., Carlini, N., Tramèr, F., & Ippolito, D. (2024). Persistent Pre-Training Poisoning of LLMs. ArXiv. https://arxiv.org/abs/2410.13722
  8. Dozono, K., Gasiba, T. E., & Stocco, A. (2024). Large Language Models for Secure Code Assessment: A Multi-Language Empirical Study. ArXiv. https://arxiv.org/abs/2408.06428
  9. Agnew, W., Jiang, H. H., Sum, C., Sap, M., & Das, S. (2024). Data Defenses Against Large Language Models. ArXiv. https://arxiv.org/abs/2410.13138.
  10. Shayegani, E., Mamun, M. A., Fu, Y., Zaree, P., & Dong, Y. (2023). Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attacks. ArXiv. https://arxiv.org/abs/2310.10844.
  11. Zheng, Z., & Zhu, X. (2023). NatLogAttack: A framework for attacking natural language inference models with natural logic. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics. Vol. 1, pp. 9960- 9976. https://doi.org/10.18653/v1/2023.acl-long.542.
  12. Shi, Y., Gao, Y., Lai, Y., Wang, H., Feng, J., He, L., Wan, J., Chen, C., Yu, Z., & Cao, X. (2024). SHIELD: An Evaluation Benchmark for Face Spoofing and Forgery Detection with Multimodal Large Language Models. ArXiv. https://arxiv.org/abs/2402.04178
  13. Zhang, J., Wang, C., Li, A., Sun, W., Zhang, C., Ma, W., & Liu, Y. (2024). An Empirical Study of Automated Vulnerability Localization with Large Language Models. ArXiv. https://arxiv.org/abs/2404.00287.
  14. Liu, Y., Gao, L., Yang, M., Xie, Y., Chen, P., Zhang, X., & Chen, W. (2024). VulDetectBench: Evaluating the Deep Capability of Vulnerability Detection with Large Language Models. ArXiv. https://arxiv.org/abs/2406.07595.
  15. Shestov, A., Levichev, R., Mussabayev, R., Maslov, E., Cheshkov, A., & Zadorozhny, P. (2024). Finetuning Large Language Models for Vulnerability Detection. ArXiv. https://arxiv.org/abs/2401.17010.
  16. Shi, Y., Gao, Y., Lai, Y., Wang, H., Feng, J., He, L., Wan, J., Chen, C., Yu, Z., & Cao, X. (2024). SHIELD: An Evaluation Benchmark for Face Spoofing and Forgery Detection with Multimodal Large Language Models. ArXiv. https://arxiv.org/abs/2402.04178