← Home
Peer-reviewed research · 2026
Runtime Authority-Aware Threat Modeling
A framework for analyzing intent, context, authority, tools, memory, delegation, and action composition in agentic AI.
When agents act, authority becomes the attack surface.
The framework examines how untrusted influence can reach privileged actions through a sequence of agent decisions. It models the authority available to an agent, influence flows through context and memory, and the graph of identities, tools, and delegated permissions.
Seven dimensionsIntent · Context · Authority · Tool capability · Memory · Delegation · Action composition
Practical questionsWhich action was requested, which authority was used, and what changed across the workflow?
Design focusIndependent authorization of consequential actions and visibility into transitive privilege.
Annam, R. (2026). Runtime Authority-Aware Threat Modeling for Agentic AI Systems: A Framework for Intent, Tool, Memory, and Delegation Risk. Journal of Recent Trends in Computer Science and Engineering, 14(4), 10–26.