Publications & Writing
Research, technical reports, and practitioner writing on AI security, cybersecurity architecture, and incident response.
13 publications
Research and technical reports
Running an Agentic Threat Model
An editorially reviewed technical article on agent identity, authority, tools, memory, and runtime threat modeling. DOI: 10.1145/3848521.
Read article →Runtime Authority-Aware Threat Modeling for Agentic AI Systems: A Framework for Intent, Tool, Memory, and Delegation Risk
Ravindra Annam · Journal of Recent Trends in Computer Science and Engineering. DOI: 10.70589/JRTCSE.2026.14.4.2.
Read paper →The AI Runtime Threat Matrix
Ravindra Annam · Version 1.0 technical report. DOI: 10.5281/zenodo.22063227.
View report →Industry articles
5 things I would never let an AI agent do without a second approval
A practical case for independent approval of consequential agent actions.
AI agents need their own identity before they need a gateway
Identity and attribution for autonomous agents.
The first 24 hours of an AI agent security incident
Containing an agent incident, preserving evidence, and recovering safely.
Practical lessons from deploying AI securely at scale
Security architecture and operational controls for enterprise AI adoption.
Accepted conference papers
Adaptive Runtime Guardrails for Preventing Prompt Injection Attacks in Agentic AI Workflows
Registered for an online presentation at ETIC, December 11–12, 2026.
View the conferenceGraph-Based Threat Detection for Tool-Using AI Agents in Enterprise Environments
The November 7 online program lists this paper, and GICITE issued an author certificate following the conference.
View the conference programAdditional research publications.
Secure Tool Invocation for LLM Agents Using Context-Aware Policy Enforcement
Ravindra Babu Annam · Context-aware policy enforcement for secure LLM-agent tool calls. DOI: 10.5281/zenodo.23204487.
Publisher record →Securing Retrieval-Augmented Generation Against Data Poisoning and Indirect Prompt Injection
Ravindra Babu Annam · Protecting retrieval-augmented generation against data poisoning and indirect prompt injection. DOI: 10.5281/zenodo.23184152.
Publisher record →Writing across audiences.
KI-Agent gehackt – ein Leitfaden für die ersten 24 Stunden
German-language edition of the AI-agent incident-response article, linked by the publisher to the original English-language work.
KI-Agent gehackt – ein Leitfaden für die ersten 24 Stunden
German-language edition of the CSO Online incident-response article, published under Ravindra Annam’s byline. This is a localized edition of the same work.