Abstract

Agentic AI systems are increasingly used across platforms for autonomous networking, security operations, observability, customer support, and business process automation. These systems consist of dynamically composed chains of AI agents, tools, and human participants that exchange highly sensitive data such as network telemetry, configuration state, customer information, and AI reasoning artifacts. Traditional security mechanisms such as static access controls, certificate-based encryption, or policy-heavy approaches like attribute-based encryption, do not scale effectively in these dynamic, high-churn environments.

The proposal introduces an Identity-Based Encryption (IBE) framework specifically designed for securing agentic AI workflows. In the proposed system, each agent, tool, and workflow execution context is assigned a cryptographic identity, and workflow data is encrypted directly to these identities. A trusted Key Generation Center issues identity-bound private keys, enabling authorization to be enforced cryptographically without certificates or complex policy evaluation.

By aligning encryption directly with agentic identities and execution semantics, the solution enables fine-grained, least-privilege data sharing, strong tenant and workflow isolation, and cryptographic containment of AI reasoning and operational data. The approach supports ephemeral, run-scoped identities, hop-by-hop re-encryption across agent chains, and efficient revocation with minimal blast radius.

The proposed architecture provides a scalable and practical security foundation for autonomous networking, zero-trust security platforms, AIOps, multi-tenant SaaS offerings, and human-in-the-loop AI systems. It reduces operational complexity, strengthens data protection guarantees, and enables safe adoption of agentic AI across different products.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS