IETF agent identity documents sorted into five maturity tiers, from published RFCs down to individual drafts

Agent Identity at the IETF: The Finished Parts Are General OAuth, the Agent Parts Are Drafts

2026-11-13 00:00:00 +0100 AI Security, Agentic AI, Identity, Standards

Authority held at each hop of an agent delegation chain, constant under bearer delegation and narrowing under attenuation

Agent Delegation Chains: The Standard Enforces Only the Last Hop

2026-10-23 00:00:00 +0200 AI Security, Agentic AI, Identity, OAuth

Six places agent and MCP flaws live, ranked by how well a CVE record describes them

Why Your CVE Program Misses Most AI Agent Vulnerabilities

2026-10-11 00:00:00 +0200 AI Security, Agentic AI, MCP, Vulnerability Management

The lethal trifecta - private data, untrusted content, and an outbound path

The Lethal Trifecta in Production: How Enterprise Copilots Quietly Became Exfiltration Tools

2026-09-22 00:00:00 +0200 AI Security, Agentic AI, Prompt Injection

Two protocol layers each threat-modeled in isolation, with the attack living in the boundary between them

The Trust Gap: Your Threat Model Stops at the Protocol Boundary

2026-09-15 00:00:00 +0200 AI Security, Agentic AI, Threat Modeling

Poisoned documents positioned closer to a query vector than the legitimate document

Cosine Similarity Is Not a Safety Property

2026-09-08 00:00:00 +0200 AI Security, RAG, Embeddings

Four injection variants escalating from structural markers to pure natural language

The Injection That Passes Every Filter

2026-09-01 00:00:00 +0200 AI Security, Prompt Injection, RAG

A five-layer redaction architecture sitting between tool execution and the model's context window

The Missing Control: A Design Proposal for Tool-Output Redaction in AI Agents

2026-08-29 00:00:00 +0200 AI Security, Agentic AI, Tool Security, Data Protection, Architecture, Cloud Security

Workload identity and delegated identity as two layers, with the authorization gap left open between them

Agent Identity: What the Standards Fix, and What They Leave Open

2026-08-22 00:00:00 +0200 AI Security, Agentic AI, Identity, SPIFFE, OAuth, Cloud Security

Bar chart comparing prompt injection attack success rates across benchmarks, showing undefended vs defended performance

What Prompt Injection Benchmarks Actually Measure

2026-08-14 00:00:00 +0200 AI Security, Agentic AI, Prompt Injection, Benchmarks, Security Research, Cloud Security

15+

Years in Production Security

5

Critical Sectors of

Professional Experience

PhD

Computer Science + CISSP