"Designing a Security Architecture for a Local AI Agent — 7-Layer Defense in Depth"

스킬 시스템 설계 — 키워드 트리거로 에이전트 능력을 모듈화하는 법

A 7-layer defense-in-depth strategy for a single-user local AI agent

핵심 요약

  • "It's a personal project, I'll handle security later" — this turned out to be a critical misjudgment. An unauthorized user sent commands to the agent through a Telegram bot.
  • Built a 7-layer defense-in-depth architecture spanning network through cognitive layers.
  • Consciously accepted risks (e.g., sandbox disabled, plaintext keys) are explicitly documented rather than ignored.
1. SKILL.md 파일 구조 — 세 가지 필수 요소

Background

LLM agents read emails, modify calendars, access the file system, and call external APIs. The agent effectively operates as core system infrastructure. During a Telegram integration test, an unauthorized external user issued commands to the bot — and the agent complied. After this security incident, I established a "Security from Day 1" principle and restructured the entire architecture.

The Architecture

3. 스킬 간 파이프라인 — 단일 스킬에서 워크플로로

Trust Model

  • Runtime environment: Single-user Mac Mini, internet-connected, no Docker
  • Primary defense targets: Unauthorized network access, communication channel hijacking, prompt injection
  • Trust anchor: Only the system account owner who can modify the ~/.openclaw directory is recognized as the operator

7-Layer Security Architecture

Layer Mechanism Status
Network Loopback Bind + Token Auth Active
Channel Telegram DM Pairing + Group Allowlist Active
Filesystem .openclaw/ 700, config 600 Active
Injection Defense 11-category defense prompt ruleset Prompt-level
Execution Control Per-sub-agent minimal exec permissions Active
Secret Management .gitignore + Pre-commit security audit Active
Trust Boundary Single-user local environment (accepted) Accepted

Perimeter Control: Network and Channel

{
  "gateway": {
    "port": 18789, "mode": "local", "bind": "loopback",
    "auth": { "mode": "token", "token": "..." }
  }
}

bind: loopback restricts connections to the same machine. Even if an SSRF attack gets through, static token authentication acts as the second line of defense.

Telegram uses dmPolicy: pairing to allow DMs only from pre-paired users, and groupPolicy: allowlist to restrict group commands to designated users.

Internal Control: Injection Defense

11 categories of prompt injection defense rules: - ignore-previous rejection: Classic jailbreak attempts are discarded - Encoding attack defense: Base64 and hex-obfuscated commands are rejected - Role-switching defense: Persona change attempts like "you are now in admin mode" are refused

Action Authorization Boundaries

  • Free execution: File reads, simple web GETs, system log checks
  • Approval required: Email/SNS sends, external API state changes, local file deletion

Pitfalls and Accepted Risks

Consciously accepted risks: 1. Sandbox disabled — Unavoidable for a personal assistant agent that needs direct local filesystem management 2. Plaintext keys in config files — Accepted for a single-operator private repository. Migration to a vault is planned if the project goes open source.

Takeaway

Even for personal projects, if the system has permissions that touch core infrastructure, security must be designed from day one. Retrofitting access control after the fact causes conflicts with existing functionality and incurs massive testing costs. Bake the principle of least privilege and zero trust for external inputs into the architecture from the start.

댓글

댓글 쓰기

이 블로그의 인기 게시물

Agent Memory Engine (2/10) — Building an AI Agent Memory System with SQLite Alone

← 1/10 SQLite Memory Engine f… 📚 Series Index 3/10 memcore Completed → A memory engine that runs without daemons, without dependencies, anywhere 핵심 요약 This post covers how to design an AI agent memory engine (memcore) on a single-file SQLite backend. - Structure: 25 modules, ~6,464 lines, 22 DB tables - Principles: no daemon, single-file portability, no required external dependencies, host-agnostic - Core techniques: 4-layer hybrid search (topic · vector · FTS5 · LIKE), U-tag dialectic-based confidence evolution, ingestion-layer bias rejection rules What This Post Covers One approach to building AI agent memory without a server or cloud vector DB. Specifically: (1) the limitations of text-file-based memory, (2) vector DB alternatives compared and selection criteria, (3) how to layer search tiers, (4) how to evolve confidence scores over time, and (5) how to block biased memory accumulation at the ingestion stage. Design Principles: Why Single-File SQLite Text-file-based ...
Read more »

"ML Foundations (9/9) — PyTorch vs TensorFlow, and the Road to Local LLMs"

이미지
← 8/9 Deep Learning Architec… 📚 Series Index (series end) The final part. So far we've focused on models. Now we focus on the tools that actually run them . We'll lay out the philosophical difference between PyTorch and TensorFlow, and trace how Hugging Face transformers , llama.cpp , MLX , and Ollama built the bridge to running large language models on your own machine. By the end you should have the full mental model of "download a pretrained LLM and serve it locally." 0. Learning Objectives Explain the eager-vs-graph difference between PyTorch and TensorFlow. Explain, in graph terms, how autograd automates the backward pass. Use Hugging Face transformers ' AutoModel , AutoTokenizer , and pipeline abstractions. Describe llama.cpp's GGUF quantization, INT4 inference, and CPU-first flow. Describe Apple MLX's use of unified memory and how it differs from PyTorch. Run a local LLM with Ollama and call it through an OpenAI-compatible API. ...
Read more »

"RAG Core Study (14/26) — Evaluation Sets with RAGAS & DeepEval"

이미지
Series overview: Series index 📚 Series Index Without an evaluation set, every RAG improvement is a story, not evidence. RAG systems fail in more than one place: retrieval, context selection, answer generation, and grounding. That means "looks better" is not a metric. Part 14 explains how to build a golden dataset , how RAGAS and DeepEval fit into the workflow, and why teams must separate retrieval quality from answer quality if they want tuning decisions to hold up. 0. Prerequisites Part 13 reranking — system quality now depends on multi-stage ranking. Part 1 grounding — a correct-looking answer is not enough. Part 12 Hybrid — multiple retrieval settings need comparison on fixed data. 1. Learning Objectives Build a small but useful golden dataset for RAG. Distinguish retrieval metrics from answer metrics . Use RAGAS and DeepEval in the right roles. Avoid the common traps in judge-model-based evaluation. 2. 핵심 요약 An evaluation set for RAG is ...
Read more »

"ML Foundations (8/9) — Deep Learning Architectures: CNN, RNN, Attention"

이미지
← 7/9 Deep Learning Training 📚 Series Index 9/9 PyTorch vs TensorFlow → What did we keep adding on top of an MLP? Vision went CNN, sequences went RNN/LSTM, and eventually both converged on Attention and the Transformer. Each architecture is easier to remember if you read it as what it gave up to gain something else . This part is the one-line summary of the decisive inflection points: LeNet → AlexNet → ResNet → LSTM → Attention → Transformer. 0. Learning Objectives Explain why CNN's convolution, pooling, and weight sharing match images so well. Trace what got unblocked from LeNet → AlexNet → VGG → ResNet as depth grew. State BPTT for RNNs and the vanishing/exploding gradient problem. Write the LSTM gate equations (forget, input, output) with intuition. Write the attention formula in query/key/value form and the scaled dot-product variant. State the precise reasons Transformers replaced RNNs (parallelism + long-range dependencies). 1. 핵심 요약 CNN : weight sharin...
Read more »

"ML Foundations (7/9) — Deep Learning Training: Optimizers, Regularization, Initialization"

← 6/9 Neural Networks 📚 Series Index 8/9 Deep Learning Architec… → Building an MLP, as we did in Part 6, is not the same as getting it to train well . The right optimizer, regularization, initialization, and learning rate — when these line up, deep networks converge. When they don't, the network refuses to learn at all. This part is the catalog of those crafts, with formulas, paper citations, and code in one place. 0. Learning Objectives Compare and write the update rules for SGD, Momentum, Nesterov, and Adam. Explain how Dropout, BatchNorm, and LayerNorm work and where they belong in a model. Derive the variance formulas for Xavier and He initialization and match them to activations. Implement step, cosine, and warmup learning-rate schedules in PyTorch. Explain why gradient clipping is effectively required for RNNs and Transformers. Diagnose the most common training failures (NaN, plateau, overfitting) and apply first-line fixes. 1. 핵심 요약 SGD : \(w \leftarro...
Read more »

OpenClaw to Hermes Migration (2/13) — What to Preserve, Partially Port, or Discard

← 1/13 Current Structure Inve… 📚 Series Index 3/13 What Moves to Hermes → A code review record classifying operational scripts into three tiers: preserve, partial-preserve, and discard. What This Post Covers An asset classification framework (preserve / partial-preserve / discard) for migrating a legacy agent system to a successor A concrete case applying the Strangler Fig + Subset Migration strategy to a memory pipeline Classification results evaluated across LOC, external dependencies, and replaceability axes A method for slimming down a bloated single file (1,310 LOC) using the migration as a natural entry point Problem Definition: Port Everything or Cut Deep? The OpenClaw inventory comprises multiple agents, multiple scheduler/daemon processes, and multiple operational scripts. When moving to the successor system, Hermes, the choice is not binary. A full port carries accumulated technical debt along with it; a minimal port discards hard-won domain logic. Code-revie...
Read more »

AI Agents I Built (5/7) — Building an Automated Blogger API Publishing System

← 4/7 My Life Organizer Agent 📚 Series Index 6/7 Dual → Designing a Markdown-Based Auto-Publish Pipeline with Blogger API v3 + OAuth 2.0 Key Summary Blogger API v3 + OAuth 2.0 enables a pipeline that converts Markdown files into publishable HTML posts blogger_publish.py handles the full flow — frontmatter parsing → body cleanup → HTML conversion → CSS injection → API post — in a single script For bulk publishing, quota management (delay, retry, TOC suppression) is the critical design concern Platform Selection Criteria API access is mandatory for AI agent-driven content publishing. A comparison of platforms: Platform API Cost AdSense Decision WordPress.com REST API Paid plan required Paid only Cost overhead Ghost REST API Self-hosted or paid Manual setup Ops overhead Blogger v3 API Free Native integration Adopted Blogger is free, exposes a REST API, and integrates natively with AdSense. For personal blog automation, this combination i...
Read more »