🛡️ OpenClaw 企业级安全配置指南

🎯 企业部署的安全矩阵

企业级 Agent 部署不是"把代码跑起来"就完事了。你需要面对的是：CISO 的灵魂拷问、审计团队的放大镜、合规认证的 checklist，以及无时无刻不在扫描你系统的黑客。

OpenClaw 的安全架构设计得像洋葱一样——层层剥离，层层防护。本文档提供完整的企业级安全配置方案，帮你构建从边界到内核的纵深防御体系。

⚙️ 安全配置详解

2.1 认证和授权配置

OpenClaw 支持多种认证方式，企业部署推荐分层认证策略：API Key 用于服务间通信，JWT 用于用户会话，OAuth2 对接企业 SSO。

# config/security/auth.yaml
auth:
  api_keys:
    enabled: true
    rotation:
      enabled: true
      interval_days: 90
    storage:
      type: "kms"
      key_id: "${KMS_KEY_ID}"
  
  jwt:
    enabled: true
    algorithm: "RS256"
    private_key_path: "/etc/openclaw/ssl/jwt_private.pem"
    access_token_ttl: 3600  # 1小时
    refresh_token_ttl: 2592000  # 30天
  
  oauth2:
    enabled: true
    provider: "okta"
    client_id: "${OAUTH_CLIENT_ID}"
    client_secret: "${OAUTH_CLIENT_SECRET}"
    authorization_url: "https://your-org.okta.com/oauth2/v1/authorize"
    auto_create_user: true
    default_role: "viewer"

rbac:
  roles:
    - name: "admin"
      permissions: ["*"]
    - name: "operator"
      permissions: ["agent:read", "agent:write", "skills:read", "skills:execute"]
    - name: "viewer"
      permissions: ["agent:read", "skills:read"]
    - name: "auditor"
      permissions: ["audit:read", "logs:read"]

2.2 Skills 权限沙箱配置

Skills 是 Agent 的能力扩展，但也是攻击面。通过 Sandbox 隔离，即使 Skill 被恶意代码污染，也能将损失控制在沙箱内。

# config/security/sandbox.yaml
sandbox:
  default_policy: "deny-all"
  
  filesystem:
    read_allowed:
      - "/usr/local/lib/openclaw/skills"
      - "/var/lib/openclaw/skills"
    write_allowed:
      - "/tmp/openclaw-${SKILL_ID}"
      - "/var/log/openclaw/skills"
    forbidden:
      - "/etc/shadow"
      - "/root"
      - "/.ssh"
  
  network:
    egress:
      default: "deny"
      allowed_hosts:
        - "api.openai.com"
        - "api.anthropic.com"
        - "*.your-company.com"
      allowed_ports: [443, 80]
      forbidden_cidrs:
        - "10.0.0.0/8"
        - "172.16.0.0/12"
        - "192.168.0.0/16"
  
  process:
    seccomp: true
    namespaces: ["pid", "net", "mnt"]
    forbidden_syscalls: ["ptrace", "mount", "setuid"]
  
  resources:
    cpu_limit: "500m"
    memory_limit: "512Mi"
    timeout_seconds: 300

2.3 数据加密配置

企业数据需要全链路加密：传输层 TLS 1.3 + 存储层 AES-256 + 密钥信封加密。

# config/security/encryption.yaml
encryption:
  tls:
    enabled: true
    min_version: "1.2"
    max_version: "1.3"
    force_https: true
    hsts:
      enabled: true
      max_age: 31536000
    ciphers_tls13:
      - "TLS_AES_256_GCM_SHA384"
  
  storage:
    database:
      enabled: true
      algorithm: "AES-256-GCM"
      kms_key_id: "${KMS_KEY_ID}"
  
  kms:
    provider: "aws-kms"
    key_rotation:
      enabled: true
      interval_days: 365

2.4 审计日志配置

审计日志是安全事件的"黑匣子"。配置要求：不可篡改、全链路覆盖、实时告警。

# config/security/audit.yaml
audit:
  level: "info"
  
  outputs:
    - type: "file"
      path: "/var/log/openclaw/audit.log"
      rotation:
        max_size_mb: 100
        max_files: 30
    - type: "syslog"
      host: "siem.your-company.com"
      port: 514
      tls: true
  
  events:
    authentication:
      log_success: true
      log_failure: true
    authorization:
      log_permission_denied: true
    agent_operations:
      log_create: true
      log_delete: true
      mask_sensitive: true
    skill_execution:
      log_sandbox_violations: true
  
  alerts:
    - name: "Multiple Auth Failures"
      condition: "authentication.failure >= 5 within 5m"
      severity: "high"
    - name: "Sandbox Violation"
      condition: "skill.sandbox_violation >= 1"
      severity: "critical"
      auto_block_skill: true
  
  integrity:
    signing:
      enabled: true
      algorithm: "RSA-SHA256"

✅ 生产环境安全检查清单

部署前，拿着这个 Checklist 逐项打勾。CEO 问起来，你心里有底。

💻 代码示例：安全配置实战

4.1 企业部署完整配置文件

生产环境推荐的最小安全配置集合：

# config/security/production.yaml
# OpenClaw 企业级安全配置 (生产环境)

# 1. 认证授权
auth:
  api_keys:
    enabled: true
    rotation:
      enabled: true
      interval_days: 90
  jwt:
    enabled: true
    algorithm: "RS256"
    access_token_ttl: 3600
  oauth2:
    enabled: true
    provider: "okta"
    auto_create_user: true

rbac:
  roles:
    - name: "admin"
      permissions: ["*"]
    - name: "operator"
      permissions: ["agent:read", "agent:write"]

# 2. 沙箱隔离
sandbox:
  default_policy: "deny-all"
  filesystem:
    read_allowed: ["/usr/local/lib/openclaw/skills"]
    write_allowed: ["/tmp/openclaw-${SKILL_ID}"]
    forbidden: ["/etc/shadow", "/root"]
  network:
    egress:
      default: "deny"
      allowed_hosts: ["api.openai.com", "*.your-company.com"]
  resources:
    cpu_limit: "500m"
    memory_limit: "512Mi"

# 3. TLS 加密
encryption:
  tls:
    enabled: true
    min_version: "1.2"
    force_https: true
    hsts:
      enabled: true
  storage:
    database:
      enabled: true
      algorithm: "AES-256-GCM"

# 4. 审计日志
audit:
  level: "info"
  outputs:
    - type: "file"
      path: "/var/log/openclaw/audit.log"
  events:
    authentication:
      log_failure: true
    skill_execution:
      log_sandbox_violations: true
  alerts:
    - name: "Sandbox Violation"
      severity: "critical"
      auto_block_skill: true

# 5. 网络策略
network:
  ingress:
    - port: 443
      protocol: "tcp"
  egress:
    default: "deny"
    allowed:
      - host: "api.openai.com"
        port: 443

# 6. CVE 扫描
scanning:
  dependencies:
    enabled: true
    tool: "trivy"
    severity_threshold: "HIGH"
    fail_on_vulnerability: true

# 7. 合规配置
compliance:
  soc2:
    enabled: true
  gdpr:
    enabled: true
    data_subject_rights:
      export_enabled: true
      delete_enabled: true

4.2 权限检查脚本 (Bash)

部署前运行此脚本，检查常见权限配置错误：

#!/bin/bash
# check-permissions.sh - OpenClaw 安全权限检查
set -e
RED='\033[0;31m'; GREEN='\033[0;32m'; YELLOW='\033[1;33m'; NC='\033[0m'
PASS=0; FAIL=0; WARN=0

echo "======== OpenClaw 安全权限检查 ========"

# 1. 检查配置文件权限
check_config_permissions() {
    echo -e "${YELLOW}[检查]${NC} 配置文件权限..."
    JWT_KEY="/etc/openclaw/ssl/jwt_private.pem"
    if [ -f "$JWT_KEY" ]; then
        PERM=$(stat -c %a "$JWT_KEY")
        if [ "$PERM" != "600" ]; then
            echo -e "  ${RED}[FAIL]${NC} JWT 私钥权限应为 600"
            FAIL=$((FAIL+1))
        else
            echo -e "  ${GREEN}[PASS]${NC} JWT 私钥权限正确"
            PASS=$((PASS+1))
        fi
    fi
}

# 2. 检查环境变量
check_env_secrets() {
    echo -e "${YELLOW}[检查]${NC} 环境变量..."
    if grep -r "sk-" /etc/openclaw/config/ 2>/dev/null | grep -v "\${.*}"; then
        echo -e "${RED}[FAIL]${NC} 发现硬编码 API Key！"
        FAIL=$((FAIL+1))
    else
        echo -e "${GREEN}[PASS]${NC} 未发现硬编码密钥"
        PASS=$((PASS+1))
    fi
}

# 3. 检查沙箱配置
check_sandbox() {
    echo -e "${YELLOW}[检查]${NC} 沙箱配置..."
    if grep -q "seccomp: true" /etc/openclaw/config/security/sandbox.yaml 2>/dev/null; then
        echo -e "${GREEN}[PASS]${NC} seccomp 已启用"
        PASS=$((PASS+1))
    else
        echo -e "${RED}[FAIL]${NC} seccomp 未启用"
        FAIL=$((FAIL+1))
    fi
}

# 4. 检查 TLS
check_tls() {
    echo -e "${YELLOW}[检查]${NC} TLS 配置..."
    if grep -q "force_https: true" /etc/openclaw/config/security/encryption.yaml 2>/dev/null; then
        echo -e "${GREEN}[PASS]${NC} 强制 HTTPS 已启用"
        PASS=$((PASS+1))
    else
        echo -e "${RED}[FAIL]${NC} 未强制 HTTPS"
        FAIL=$((FAIL+1))
    fi
}

check_config_permissions
check_env_secrets
check_sandbox
check_tls

echo ""
echo "======== 检查结果 ========"
echo -e "${GREEN}通过: $PASS${NC}"
echo -e "${RED}失败: $FAIL${NC}"
echo -e "${YELLOW}警告: $WARN${NC}"

if [ $FAIL -gt 0 ]; then
    echo -e "${RED}❌ 安全检查失败，

    echo -e "${GREEN}[PASS]${NC} 强制 HTTPS 已启用"
        PASS=$((PASS+1))
    fi
}

check_config_permissions
check_env_secrets
check_sandbox
check_tls

echo ""
echo "======== 检查结果 ========"
echo -e "${GREEN}通过: $PASS${NC}"
echo -e "${RED}失败: $FAIL${NC}"
echo -e "${YELLOW}警告: $WARN${NC}"

if [ $FAIL -gt 0 ]; then
    echo -e "${RED}❌ 安全检查失败，请修复后重新部署${NC}"
    exit 1
else
    echo -e "${GREEN}✅ 所有安全检查通过${NC}"
    exit 0
fi