# Firewalls to Secure Dynamic LLM Agentic Networks

arXiv: https://arxiv.org/abs/2502.01822 · HTML: https://mlsanity.com/paper/2502.01822/

## Key facts

- Cited by 36 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 19 of its 60 references that are in the map.
- First posted on arXiv 3 Feb 2025.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 19 of 19)

- [Membership Inference Attacks against Machine Learning Models](https://mlsanity.com/paper/1610.05820/) (2016)
- [Extracting Training Data from Large Language Models](https://mlsanity.com/paper/2012.07805/) (2020)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://mlsanity.com/paper/2302.12173/) (2023)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [Identifying the Risks of LM Agents with an LM-Emulated Sandbox](https://mlsanity.com/paper/2309.15817/) (2023)
- [TravelPlanner: A Benchmark for Real-World Planning with Language Agents](https://mlsanity.com/paper/2402.01622/) (2024)
- [The Composition Theorem for Differential Privacy](https://mlsanity.com/paper/1311.0776/) (2013)
- [Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory](https://mlsanity.com/paper/2310.17884/) (2023)
- [PrivacyLens: Evaluating Privacy Norm Awareness of Language Models in Action](https://mlsanity.com/paper/2409.00138/) (2024)
- [AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents](https://mlsanity.com/paper/2406.13352/) (2024)
- [Cooperation, Competition, and Maliciousness: LLM-Stakeholders Interactive Negotiation](https://mlsanity.com/paper/2309.17234/) (2023)
- [System-Level Defense against Indirect Prompt Injection Attacks: An Information Flow Control Perspective](https://mlsanity.com/paper/2409.19091/) (2024)
- [AirGapAgent: Protecting Privacy-Conscious Conversational Agents](https://mlsanity.com/paper/2405.05175/) (2024)
- [Get my drift? Catching LLM Task Drift with Activation Deltas](https://mlsanity.com/paper/2406.00799/) (2024)
- [Contextual Agent Security: A Policy for Every Purpose](https://mlsanity.com/paper/2501.17070/) (2025)
- [A Scalable Communication Protocol for Networks of Large Language Models](https://mlsanity.com/paper/2410.11905/) (2024)
- [Adversarial Search Engine Optimization for Large Language Models](https://mlsanity.com/paper/2406.18382/) (2024)
- [Operationalizing Contextual Integrity in Privacy-Conscious Assistants](https://mlsanity.com/paper/2408.02373/) (2024)
- [Dataset and Lessons Learned from the 2024 SaTML LLM Capture-the-Flag Competition](https://mlsanity.com/paper/2406.07954/) (2024)

## Built on by (most cited first, 30 of 36)

- [Defeating Prompt Injections by Design](https://mlsanity.com/paper/2503.18813/) (2025)
- [Large Language Model Agent: A Survey on Methodology, Applications and Challenges](https://mlsanity.com/paper/2503.21460/) (2025)
- [A survey of agent interoperability protocols: Model Context Protocol (MCP), Agent Communication Protocol (ACP), Agent-to-Agent Protocol (A2A), and Agent Network Protocol (ANP)](https://mlsanity.com/paper/2505.02279/) (2025)
- [OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety](https://mlsanity.com/paper/2507.06134/) (2025)
- [Design Patterns for Securing LLM Agents against Prompt Injections](https://mlsanity.com/paper/2506.08837/) (2025)
- [Open Challenges in Multi-Agent Security: Towards Secure Systems of Interacting AI Agents](https://mlsanity.com/paper/2505.02077/) (2025)
- [MCPSecBench: A Systematic Security Benchmark and Playground for Testing Model Context Protocols](https://mlsanity.com/paper/2508.13220/) (2025)
- [DoomArena: A framework for Testing AI Agents Against Evolving Security Threats](https://mlsanity.com/paper/2504.14064/) (2025)
- [Contextual Integrity in LLMs via Reasoning and Reinforcement Learning](https://mlsanity.com/paper/2506.04245/) (2025)
- [LLM Agents Should Employ Security Principles](https://mlsanity.com/paper/2505.24019/) (2025)
- [The Attack and Defense Landscape of Agentic AI: A Comprehensive Survey](https://mlsanity.com/paper/2603.11088/) (2026)
- [Searching for Privacy Risks in LLM Agents via Simulation](https://mlsanity.com/paper/2508.10880/) (2025)
- [Indirect Prompt Injections: Are Firewalls All You Need, or Stronger Benchmarks?](https://mlsanity.com/paper/2510.05244/) (2025)
- [CaMeLs Can Use Computers Too: System-level Security for Computer Use Agents](https://mlsanity.com/paper/2601.09923/) (2026)
- [Breaking and Fixing Defenses Against Control-Flow Hijacking in Multi-Agent Systems](https://mlsanity.com/paper/2510.17276/) (2025)
- [Terrarium: Revisiting the Blackboard for Multi-Agent Safety, Privacy, and Security Studies](https://mlsanity.com/paper/2510.14312/) (2025)
- [The Promptware Kill Chain: How Prompt Injections Gradually Evolved Into a Multistep Malware Delivery Mechanism](https://mlsanity.com/paper/2601.09625/) (2026)
- [AI Agents May Always Fall for Prompt Injections](https://mlsanity.com/paper/2605.17634/) (2026)
- [Contextualized Privacy Defense for LLM Agents](https://mlsanity.com/paper/2603.02983/) (2026)
- [MemPrivacy: Privacy-Preserving Personalized Memory Management for Edge-Cloud Agents](https://mlsanity.com/paper/2605.09530/) (2026)
- [Toward Secure LLM Agents: Threat Surfaces, Attacks, Defenses, and Evaluation](https://mlsanity.com/paper/2606.10749/) (2026)
- [LAMeTA: Intent-Aware Agentic Network Optimization via a Large AI Model-Empowered Two-Stage Approach](https://mlsanity.com/paper/2505.12247/) (2025)
- [Towards Pervasive Distributed Agentic Generative AI -- A State of The Art](https://mlsanity.com/paper/2506.13324/) (2025)
- [Enforcing Benign Trajectories: A Behavioral Firewall for Structured-Workflow AI Agents](https://mlsanity.com/paper/2604.26274/) (2026)
- [VeriGrey: Greybox Agent Validation](https://mlsanity.com/paper/2603.17639/) (2026)
- [Alignment Contracts for Agentic Security Systems](https://mlsanity.com/paper/2605.00081/) (2026)
- [It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs](https://mlsanity.com/paper/2605.20258/) (2026)
- [SovereignNegotiation-Bench: Evaluating User-Owned Personal Agents In Delegated Bargaining Under Privacy, Consent, Evidence, And Institutional Pressure](https://mlsanity.com/paper/2607.02814/) (2026)
- [REDDIT: Forgetting-Resistant Correction of Timestamp Drift in ASR via Replay-Based Distribution Editing](https://mlsanity.com/paper/2607.05364/) (2026)
- [The Balkanization of Execution-Security Research for AI Coding Agents: Isolation, Access Control, and Time-of-Check-to-Time-of-Use Vulnerabilities](https://mlsanity.com/paper/2607.05743/) (2026)

## Abstract (arXiv)

The emergence of agent-to-agent communication protocols mirrors the early internet: powerful connectivity with minimal security infrastructure. When AI agents communicate on behalf of users, every message crosses a trust boundary where the user's personal data and the external agent's unconstrained language each present distinct risks. We address both through a dual-firewall architecture grounded in a unifying principle: each task defines a context, and both sides of the communication carry information far exceeding what that context requires. Our firewalls act as projections onto the task context, allowing only contextually appropriate content to cross each boundary. The Language Converter Firewall projects incoming messages onto a closed, domain-specific, structured protocol; an external agent's message is converted to validated fields while persuasive framing, urgency tactics, and embedded instructions are structurally eliminated through deterministic verification. This replaces the asymmetric challenge of resisting every possible manipulation with the structural guarantee that manipulation has no channel through which to arrive. The Data Abstraction Firewall projects outgoing information onto the granularity appropriate for the task, rather than applying binary disclose-or-redact filtering, as previous airgapping solutions did. Both firewalls operate in a trusted environment isolated from external input, applying domain-specific rules learned automatically from demonstrations. Across 864 attacks spanning three domains on the ConVerse benchmark, our architecture reduces privacy attack success rates (e.g., from 84% to 10% for GPT-5) and security attacks (from 60% to 3%), while maintaining or even improving task completion quality.