A Conceptual Review of LLM-Based On-Chain AI Agents for DeFi Asset Management

Authors

  • Ridwan Yusuf Universitas Dharmawacana Metro, Metro, Indonesia Author
  • Andreas Perdana✉ Universitas Dharmawacana Metro, Metro, Indonesia Author
  • Febri Sugandi Universitas Dharmawacana Metro, Metro, Indonesia Author
  • Untoro Apsiswanto Universitas Dharmawacana Metro, Metro, Indonesia Author
Diajukan 25 Sep 2025 Diterbitkan 25 Des 2025
DOI: https://doi.org/10.59025/ef9b943c

Keywords:

AI agent, large language model, smart contract, DeFi, autonomous agent security

Abstract

Over 2024 and 2025, the encounter between Large Language Models and blockchain technology produced a new class of software called on-chain AI Agents, autonomous programs that translate user intent into chains of Decentralized Finance transactions and sign them without human intervention. The rapid adoption has exposed three conceptual gaps that remain unresolved: the complexity of multi-step DeFi execution flows, an LLM-specific vulnerability landscape such as indirect prompt injection and jailbreak that directly threatens financial assets, and the absence of an accountable governance framework within decentralised ecosystems. This study synthesises 34 verified publications from a Zotero corpus across four interconnected themes: agent architecture and multi-agent collaboration, key management within agentic wallets, the security threat landscape, and blockchain role as a trust layer. The key finding that emerges is that defences relying solely on the language layer prove insufficient; action-layer controls, through dynamic policies, enclave isolation, or multi-party secret sharing, become preconditions when the financial consequences are permanent. The contributions offered include a per-layer threat taxonomy, a three-layer hybrid defence framework, and a future research agenda with five concrete directions.

References

  1. F. Chen, "Multi-Agent LLM Systems: From Emergent Collaboration to Structured Collective Intelligence," 2025.
  2. V. Kulpa et al., "Prompt Engineering in Finance: An LLM-Based Multi-Agent Architecture for Decision Support," 2025.
  3. Q. Zhan, R. Fang, H. S. Panchal, dan D. Kang, "Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents," 2025.
  4. S. Johnson, V. Pham, dan T. Le, "The Dangers of Indirect Prompt Injection Attacks on LLM-based Autonomous Web Navigation Agents: A Demonstration," 2025.
  5. R. Shahroz, Z. Tan, S. Yun, C. Fleming, T. Chen et al., "Agents Under Siege: Breaking Pragmatic Multi-Agent LLM Systems with Optimized Prompt Attacks," 2025.
  6. Z. Liu et al., "Combining Graph Neural Networks With Expert Knowledge for Smart Contract Vulnerability Detection," IEEE Transactions on Knowledge and Data Engineering, 2023, doi: 10.1109/TKDE.2021.3095196.
  7. S. Mane, J. Bhortake, V. Wankhade, dan F. Kazi, "Automated IoT Firmware Vulnerability Detection Using Large Language Models," Preprints, 2025, doi: 10.20944/preprints202510.0166.v1.
  8. S. Fathima et al., "LLM Adversarial Prompt Attack Detection and Mitigation Engine: A Novel Framework for Securing Generative AI Systems," 2025.
  9. Q. Lan, A. Kaul, dan S. Jones, "Prompt Injection Detection in LLM Integrated Applications," 2025.
  10. D. Naik, N. Naik, dan I. Naik, "Insecure Output Handling in Large Language Models (LLMs) and Approaches to Enhance Output Security, Including Prevention of LLM-Based Attacks," 2025.
  11. S. Choudhary, D. Anshumaan, N. Palumbo, dan S. Jha, "How Not to Detect Prompt Injections with an LLM," 2025.
  12. A. Ghafarollahi dan M. J. Buehler, "Rapid and automated alloy design with graph neural network-powered large language model-driven multi-agent AI," 2025.
  13. X. Lan, J. Feng, Y. Liu, X. Shi, dan Y. Li, "AutoQual: An LLM Agent for Automated Discovery of Interpretable Features for Review Quality Assessment," 2025.
  14. Y. Teplyi dan D. Dosyn, "Prompt-guided LLM agent for end-to-end ontology learning," 2025.
  15. S. Kakkar, "Prompt-Robotics: Leveraging Large Language Model (LLM) Agents to Enhance Robotic Process Automation (RPA) Supervision and Efficiency," 2025.
  16. A. Hazime et al., "Evaluation of LLM Powered Agentic AI for Solving Multi-Arm Bandit Problems," 2025.
  17. A. Mukherjee, "Use of Agentic AI with LLM and Prompt Engineering and State-of-the-Art Machine Learning Algorithm to Detect Patterns in IoT Devices," 2025.
  18. Z. Wang et al., "AGENTVIGIL: Automatic Black-Box Red-teaming for Indirect Prompt Injection against LLM Agents," 2025.
  19. F. Jia, T. Wu, X. Qin, dan A. Squicciarini, "The Task Shield: Enforcing Task Alignment to Defend Against Indirect Prompt Injection in LLM Agents," 2025.
  20. H. An, J. Zhang, T. Du, C. Zhou, dan Q. Li, "IPIGuard: A Novel Tool Dependency Graph-Based Defense Against Indirect Prompt Injection in LLM Agents," 2025.
  21. J. Hu, J. Zhang, dan Z. Li, "Breaking Barriers: Multi-Agent Prompt Fusion for Automated LLM Jailbreaks," 2025.
  22. H. Wang et al., "DiffusionAttacker: Diffusion-Driven Prompt Manipulation for LLM Jailbreak," 2025.
  23. G. Deng, Y. Liu, Y. Li, K. Wang, Y. Zhang, dan Z. Liu, "MASTERKEY: Automated Jailbreaking of Large Language Model Chatbots," 2024.
  24. P. Reddy dan A. S. Gujral, "EchoLeak: The First Real-World Zero-Click Prompt Injection Exploit in a Production LLM System," 2025.
  25. G. Yang, Z. Sun, dan Y. Wang, "ShellBox: Adversarially Enhanced LLM-Interactive Honeypot Framework," 2025.
  26. S. Panterino dan M. Fellington, "Dynamic Moving Target Defense for Mitigating Targeted LLM Prompt Injection," 2024.
  27. A. Kavian, M. M. P. Kallehbasti, S. Kazemi, dan F. Khansari, "LLM Security Guard for Code," 2024.
  28. J. Wu et al., "Who Are the Phishers? Phishing Scam Detection on Ethereum via Network Embedding," IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2022, doi: 10.1109/TSMC.2020.3016821.
  29. J. Arcenegui, R. Arjona, R. Roman, dan I. Baturone, "Secure Combination of IoT and Blockchain by Physically Binding IoT Devices to Smart Non-Fungible Tokens Using PUFs," Sensors, vol. 21, no. 9, hal. 3119, 2021, doi: 10.3390/s21093119.
  30. N. Karandikar, A. Chakravorty, dan C. Rong, "Blockchain Based Transaction System with Fungible and Non-Fungible Tokens for a Community-Based Energy Infrastructure," Sensors, vol. 21, no. 11, hal. 3822, 2021, doi: 10.3390/s21113822.
  31. L. Ante, "The Non-Fungible Token (NFT) Market and Its Relationship with Bitcoin and Ethereum," FinTech, 2022, doi: 10.3390/fintech1030017.
  32. Y. Wang, B. M. Lucey, S. A. Vigne, dan L. Yarovaya, "Bubbles all the way down? Detecting and date-stamping bubble behaviours in NFT and DeFi markets," Journal of Chinese Economic and Business Studies, 2022, doi: 10.1080/14765284.2022.2138161.
  33. K. B. Wilson, A. Karg, dan H. Ghaderi, "Prospecting Non-Fungible Tokens in the Digital Economy: Stakeholders and Ecosystem, Risk and Opportunity," Business Horizons, 2022, doi: 10.1016/j.bushor.2021.10.007.
  34. G. D. Ritterbusch dan M. R. Teichmann, "Defining the Metaverse: A Systematic Literature Review," IEEE Access, 2023, doi: 10.1109/ACCESS.2023.3241809.
Cover JTDS Vol 3 No 2 (2025)

Downloads

How to Cite

A Conceptual Review of LLM-Based On-Chain AI Agents for DeFi Asset Management. (2025). Journal of Technology and Data Science, 3(2), 216-228. https://doi.org/10.59025/ef9b943c