AI Security Research Portal

Home Browse Wiki Index Research Questions Synthesis

Search the portal

Synthesis

Current Synthesis Raw Corpus Synthesis 2026-07-02

Research Questions

Research Questions Index RQ-20260702-001-agent-security RQ-20260702-002-benchmark-validity RQ-20260702-003-defense-generalization RQ-20260702-004-agent-protocol-security RQ-20260702-005-memory-poisoning-defense RQ-20260702-006-benchmark-to-incident-validity RQ-20260702-007-action-scoped-authorization RQ-20260702-008-rag-poisoning-controls RQ-20260702-009-ai-soc-human-factors RQ-20260702-010-agent-runtime-monitoring RQ-20260703-011-open-weight-ai-soc-evaluation

Concepts

Agent Identity and Authorization Agent Security and Tool Abuse AI Cybersecurity Operations AI Security Governance and Standards AI Security Taxonomy Concepts Index Evaluation Benchmarks for AI Security MCP and Agent Protocol Security Memory Poisoning and Agent State Model Extraction and Privacy Leakage Open Weight Models for AI SOC Prompt Injection and Context Attacks RAG and Retrieval Security Threat Models

Claims

Agent Authorization Should Be Action Scoped Agentic Systems Expand The Security Boundary Benchmarks May Not Predict Deployment Risk Claims Index Open Weight SOC Models Need Evaluation Contracts Persistent Memory Creates Poisoning And Provenance Risks Prompt Injection Defenses Depend On Deployment Context

Methods

Benchmark-Based Security Evaluation Evidence Grading for AI Security Methods Index Red Teaming Agentic AI Runtime Monitoring and Agent Gateways SOC Evaluation Parser Audit Threat Modeling Agentic Systems

Sources

\"What Happens Locally, Leaks Globally\": Detecting Privacy Leakage Risks in MCP Servers A Layered Security Framework Against Prompt Injection in RAG-Based Chatbots A Practical Guide for Securely Using Third-Party MCP Servers A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence A Survey on Model Extraction Attacks and Defenses for Large Language Models A user-centric machine learning framework for cyber security operations center Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations AECR: Automatic attack technique intelligence extraction based on fine-tuned large language model Agent Security Bench (ASB): Formalizing and Benchmarking Attacks and Defenses in LLM Agents Agent Security Bench (ASB): Formalizing and Benchmarking Attacks and Defenses in LLM-based Agents Agent-Assisted Side-Channel Attacks on Non-Prefix KV Cache in RAG AgentCanary: A Security Evaluation Framework for Autonomous AI Agents in Real Executable Environments AgentCyberRange: Benchmarking Frontier AI Systems in Realistic Cyber Ranges AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents AgentDyn: Are Your Agent Security Defenses Deployable in Real-World Dynamic Environments?AgentEvolver: Towards Efficient Self-Evolving Agent System Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges AgenticCyOps: Securing Multi-Agentic AI Integration in Enterprise Cyber Operations+ 248 more in search

Portal

AI Security Research Portal Entities Index index log

Claimsoutputactive2026-07-04ai-securityclaims

Claims Index

Use claim pages when evidence needs to be tracked, contested, or updated.

Active Claims

Agent Authorization Should Be Action Scoped
Agentic Systems Expand The Security Boundary
Benchmarks May Not Predict Deployment Risk
Open Weight SOC Models Need Evaluation Contracts
Persistent Memory Creates Poisoning And Provenance Risks
Prompt Injection Defenses Depend On Deployment Context