Secure Workload Identity and Mutual Authentication for Agentic AI Systems: A SPIFFE-Based Framework
Keywords:
Agentic AI, Workload Identity, SPIFFE, Mutual AuthenticationAbstract
The Agentic artificial intelligence systems increasingly operate as collections of autonomous or semi-autonomous software agents that interact with models, tools, services, data repositories, and other agents. This distributed operating model introduces a security problem that conventional user-centric authentication mechanisms do not adequately address: each computational workload must establish a verifiable identity and authenticate communicating workloads without relying exclusively on static credentials. This paper develops a conceptual SPIFFE-based framework for secure workload identity and mutual authentication in agentic AI ecosystems. The framework combines workload identity, cryptographic attestation, short-lived credentials, mutual authentication, authorization context, and continuous identity validation into a unified security architecture. The theoretical foundation is derived from the supplied literature on reinforcement learning, multi-agent systems, neuro-symbolic reasoning, autonomous decision-making, and security-oriented learning. These studies demonstrate that increasingly autonomous systems require mechanisms capable of supporting dynamic interactions, distributed decision processes, and changing operational contexts. Building on this foundation, the paper analyzes how SPIFFE-oriented workload identities can provide an identity abstraction for AI agents independently of their location, deployment platform, or network address. The proposed framework is evaluated conceptually through security requirements, architectural components, threat considerations, and representative agent-to-agent interaction scenarios. The findings indicate that workload-oriented identity can reduce credential persistence, improve authentication consistency, and establish a stronger trust boundary for agentic AI systems. However, identity alone does not guarantee trustworthy agent behavior; authorization, policy evaluation, model integrity, provenance, and contextual risk assessment remain necessary. The paper therefore positions SPIFFE-based identity as a foundational security layer rather than a complete security solution for agentic AI.
References
K. Arulkumaran, M. P. Deisenroth, M. Brundage, and A. A. Bharath, “Deep reinforcement learning: A brief survey,” IEEE Signal Process. Mag., vol. 34, no. 6, pp. 26–38, Nov. 2017.
T. R. Besold, “Neural-symbolic learning and reasoning: A survey and interpretation,” 2017, arXiv:1711.03902.
D. Bouneffouf and C. C. Aggarwal, “Survey on applications of neurosymbolic artificial intelligence,” 2022, arXiv:2209.12618.
L. Buşoniu, R. Babuška, and B. De Schutter, “Multi-agent reinforcement learning: An overview,” in Innovations in Multi-Agent Systems and Applications-1. Berlin, Germany : Springer, 2010, pp. 183–221.
L. Canese, “Multi-agent reinforcement learning: A review of challenges and applications,” Appl. Sci., vol. 11, no. 11, 2021, Art. no. 4948.
J. García and F. Fernández, “A comprehensive survey on safe reinforcement learning,” J. Mach. Learn. Res., vol. 16, no. 1, pp. 1437–1480, 2015.
S. Gronauer and K. Dieopold, “Multi-agent deep reinforcement learning: A survey,” Artif. Intell. Rev., vol. 55, pp. 895–943, Feb. 2022.
L. P. Kaelbling, M. L. Littman, and A. W. Moore, “Reinforcement learning: A survey,” J. Artif. Intell. Res., vol. 4, pp. 237–285, 1996.
J. Kober, J. A. Bagnell, and J. Peters, “Reinforcement learning in robotics: A survey,” Int. J. Robot. Res., vol. 32, no. 11, pp. 1238–1274, 2013.
Y. Li, “Deep reinforcement learning: An overview,” 2017, arXiv:1701.07274.
R.-Z. Liu, W. Wang, Y. Shen, Z. Li, Y. Yu, and T. Lu, “An introduction of mini-alphastar,” 2021, arXiv:2104.06890.
J. Luketina, “A survey of reinforcement learning informed by natural language,” 2019, arXiv:1906.03926.
S. Padakandla, “A survey of reinforcement learning algorithms for dynamically varying environments,” ACM Comput. Surv., vol. 54, no. 6, pp. 1–25, 2021.
K. Pappu, B. Bhushan and A. Mittal, "SPIFFE-Based Zero-Trust Authentication for AI Agent Ecosystems," 2025 International Conference on Computer and Applications (ICCA), Bahrain, Bahrain, 2025, pp. 1-7, doi: 10.1109/ICCA66035.2025.11431026.
Y. Qian, J. Wu, R. Wang, F. Zhu, and W. Zhang, “Survey on reinforcement learning applications in communication networks,” J. Commun. Inf. Netw., vol. 4, no. 2, pp. 30–39, 2019.
D. Silver, “Mastering the game of go with deep neural networks and tree search,” Nature, vol. 529, no. 7587, pp. 484–489, 2016.
V. Uc-Cetina, N. Navarro-Guerrero, A. Martin-Gonzalez, C. Weber, and S. Wermter, “Survey on reinforcement learning for language processing,” Artif. Intell. Rev., vol. 56, no. 2, pp. 1543–1575, 2023.
A. Uprety and D. B. Rawat, “Reinforcement learning for IoT security: A comprehensive survey,” IEEE Internet Things J., vol. 8, no. 11, pp. 8693–8706, Jun. 2021.
W. Wang and Y. Yang, “Towards data-and knowledge-driven artificial intelligence: A survey on neuro-symbolic computing,” 2022, arXiv:2210.15889.
D. Yu, B. Yang, D. Liu, and H. Wang, “A survey on neural-symbolic systems,” 2021, arXiv:2111.08164.
Y. Zeng, R. Cai, F. Sun, L. Huang, and Z. Hao, “A survey on causal reinforcement learning,” 2023, arXiv:2302.05209.
