"Design and Implementation of a Multi-Layered Framework for Detection and Defense against Logical Attacks in Chatbots: Enhancing the Security and Resilience of Large Language Models"

Vanitaben Hasmukhbhai Suthar; Dr. Swity Maniyar; Vanitaben Pragneshkumar Mistry

doi:10.69980/redvet.v25i1.2487

Vanitaben Hasmukhbhai Suthar
Dr. Swity Maniyar
Vanitaben Pragneshkumar Mistry

DOI: https://doi.org/10.69980/redvet.v25i1.2487

Keywords: ChatGPT Security, Logical Attacks, Chatbot Vulnerabilities, Prompt Injection, Jailbreak Attacks, Context Manipulation, Large Language Models, Adversarial Attacks, Graph Neural Network, Explainable AI, AI Security Framework, Robustness Enhancement.

Abstract

The rapid adoption of Large Language Model (LLM)-based chatbots, especially ChatGPT, has introduced significant security challenges due to their vulnerability to logical attacks such as prompt injection, jailbreak attempts, contradiction-based queries, and context manipulation. Existing security approaches mainly rely on keyword filtering and rule-based detection methods, which are insufficient for identifying complex reasoning-based attacks and providing adaptive defense mechanisms. This research proposes a multi-layered AI security framework for detecting and defending against logical attacks in ChatGPT-based chatbots. The proposed framework integrates transformer-based semantic analysis, Graph Neural Networks (GNN) for logical reasoning, Contrastive Logical Feature Optimization (CLFO) for improving adversarial feature separation, and Explainable AI (XAI) techniques including LIME, SHAP, and attention visualization for transparent decision-making. The methodology incorporates multi-turn conversational context analysis, hybrid attack classification, dynamic prompt sanitization, adversarial prompt rewriting, and response validation mechanisms. The framework is evaluated using chatbot security datasets containing normal and adversarial prompts across multiple attack categories, with performance measured using accuracy, precision, recall, F1-score, Attack Success Rate Reduction (ASRR), and Defense Effectiveness Score (DES). The proposed approach aims to enhance ChatGPT’s robustness, reliability, and trustworthiness by providing an intelligent, explainable, and adaptive defense mechanism against evolving logical attacks.

Author Biographies

Vanitaben Hasmukhbhai Suthar

PhD Scholar, Swaminarayan University, Kalol, Gujarat.

Dr. Swity Maniyar

Associate Professor – GTU, Phd Guide – Swaminarayan University, Kalol, Gujarat.

Vanitaben Pragneshkumar Mistry

Assistant Professor, BCA Department- Sardar Vallabhbhai Global University, Ahmedabad, Gujarat.

References

1. George, A. Shaji. "Beyond the Interface and How Conversational AI Is Reshaping the Future of Human-Computer Interaction." Partners Universal International Innovation Journal 3, no. 4 1-13.
2. Pooja, S., G. Gokul, K. Linkesh Mani, A. S. Raj Kumar, and R. Amutha Bharathi. "Context-Aware AI Chatbot Using Transformer-Based Models for Intelligent User Interactions."
3. Hasal, Martin, Jana Nowaková, Khalifa Ahmed Saghair, Hussam Abdulla, Václav Snášel, and Lidia Ogiela. "Chatbots: Security, privacy, data protection, and social aspects." Concurrency and Computation: Practice and Experience 33, no. 19 : e6426.
4. Gulyamov, Saidakhror, Said Gulyamov, Andrey Rodionov, Rustam Khursanov, Kambariddin Mekhmonov, Djakhongir Babaev, and Akmaljon Rakhimjonov. "Prompt Injection Attacks in Large Language Models and AI Agent Systems: A Comprehensive Review of Vulnerabilities, Attack Vectors, and Defense Mechanisms." Information 17, no. 1: 54.
5. Aslan, Ömer, Semih Serkant Aktuğ, Merve Ozkan-Okay, Abdullah Asim Yilmaz, and Erdal Akin. "A comprehensive review of cyber security vulnerabilities, threats, attacks, and solutions." Electronics 12, no. 6 : 1333.
6. Quffa, Abdallah, and Samy S. Abu-Naser. "A Rule-Based Expert System for Cybersecurity Threat Detection: Evolution, Applications, and the Hybrid AI Paradigm."
7. https://www.getastra.com/blog/ai-security/prompt-injection-attacks/
8. Hassija, Vikas, Vinay Chamola, Atmesh Mahapatra, Abhinandan Singal, Divyansh Goel, Kaizhu Huang, Simone Scardapane, Indro Spinelli, Mufti Mahmud, and Amir Hussain. "Interpreting black-box models: a review on explainable artificial intelligence." Cognitive Computation 16, no. 1: 45-74.
9. Polemi, Nineta, Isabel Praça, Kitty Kioskli, and Adrien Bécue. "Challenges and efforts in managing AI trustworthiness risks: a state of knowledge." Frontiers in Big Data 7: 1381163.
10. KAMATA, Mitsunobu. "Security Design and Evaluation of a Citi-zen-Oriented Generative AI Chatbot Using RAG (Retrieval-Augmented Generation)." International Journal of ICT Application Research 3, no. 1: 1-8.
11. Anghel, Catalin, Marian Viorel Craciun, Adina Cocu, Andreea Alexandra Anghel, Antonio Stefan Balau, Adrian Istrate, and Aurelian-Dumitrache Anghele. "EvalHack: Answer-Side Prompt Injection for Probing LLM Exam-Grading Panel Stability." Information 17, no. 3: 297.
12. Zhang, Yihao, Zeming Wei, Xiaokun Luan, Chengcan Wu, Zhixin Zhang, Jiangrong Wu, Haolin Wu, Huanran Chen, Jun Sun, and Meng Sun. "ClawWorm: Self-Propagating Attacks Across LLM Agent Ecosystems." arXiv preprint arXiv:2603.15727 .
13. ŞAŞAL, A., and AB CAN. "Prompt Injection Attacks on Large Language Models: Multi-Model Security Analysis with Categorized Attack Types.".
14. Emekci, Hakan, and Gülsüm Budakoglu. "Securing With Dual-LLM Architecture: ChatTEDU an Open Access Chatbot's Defense." IEEE Access .
15. Yi, Jingwei, Yueqi Xie, Bin Zhu, Emre Kiciman, Guangzhong Sun, Xing Xie, and Fangzhao Wu. "Benchmarking and defending against indirect prompt injection attacks on large language models." In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Vol. 1, pp. 1809-1820.
16. Azooz, Hasan Jameel. "Comprehensive, context-aware, multi-layered security framework for mitigating prompt injection attacks in large language models."
17. Liu, Xiaogeng, Zhiyuan Yu, Yizhe Zhang, Ning Zhang, and Chaowei Xiao. "Automatic and universal prompt injection attacks against large language models." arXiv preprint arXiv:2403.04957 .
18. Derner, Erik, Kristina Batistič, Jan Zahálka, and Robert Babuška. "A security risk taxonomy for prompt-based interaction with large language models." IEEE Access 12: 126176-126187.
19. Jalali, Nasir Ahmad, and Chen Hongsong. "Comprehensive framework for implementing blockchain-enabled federated learning and full homomorphic encryption for chatbot security system." Cluster Computing 27, no. 8 : 10859-10882.
20. Suárez-Varela, José, Paul Almasan, Miquel Ferriol-Galmés, Krzysztof Rusek, Fabien Geyer, Xiangle Cheng, Xiang Shi et al. "Graph neural networks for communication networks: Context, use cases and opportunities." IEEE Network 37, no. 3: 146-153.
21. Sharma, Amit, Ashutosh Sharma, Alexey Tselykh, Alexander Bozhenyuk, and Byung‐Gyu Kim. "Image and video analysis using graph neural network for Internet of Medical Things and computer vision applications." CAAI Transactions on Intelligence Te.
22. Ju, Wei, Yifan Wang, Yifang Qin, Zhengyang Mao, Zhiping Xiao, Junyu Luo, Junwei Yang et al. "Towards graph contrastive learning: A survey and beyond." arXiv preprint arXiv:2405.11868 .
23. Rane, Nitin, Saurabh Choudhary, and Jayesh Rane. "Explainable Artificial Intelligence (XAI) approaches for transparency and accountability in financial decision-making." Available at SSRN 4640316.
24. Zhang, Changsheng, and Linjun Liu. "Machine learning prediction model for medical environment comfort based on SHAP and LIME interpretability analysis." Scientific Reports 15, no. 1 : 39269.
25. Liu, Ying, Yating Fu, Yadong Peng, and Jie Ming. "Clinical decision support tool for breast cancer recurrence prediction using SHAP value in cooperative game theory." Heliyon 10, no. 2 .
26. Soydaner, Derya. "Attention mechanism in neural networks: where it comes and where it goes." Neural Computing and Applications 34, no. 16: 13371-13385.