Echo
eba8849986
Completed comprehensive spec-driven documentation: - SPEC.md (222 lines): Project scope, objectives, constraints - ARCHITECTURE.md (290 lines): System design, components, data flow - REQUIREMENTS.md (168 lines): Functional & non-functional requirements - API_SPEC.md (556 lines): 15 API endpoints with schemas - SECURITY.md (188 lines): STRIDE threat model, security controls - ROADMAP.md (203 lines): 5 phases, 8 milestones, risk register Total: 1,627 lines of specification documentation Milestone M1 complete - Ready for Phase 0 (Rust scaffolding)
190 lines
7.5 KiB
Markdown
190 lines
7.5 KiB
Markdown
# Linux_Patch_API - Security Specification Document
|
|
|
|
## Security Overview
|
|
|
|
**Philosophy:** Defense in depth with zero-trust principles for internal network.
|
|
|
|
**Approach:**
|
|
- mTLS certificate-based authentication (required for all connections)
|
|
- IP whitelist enforcement (deny by default, allow only listed)
|
|
- Comprehensive audit logging for all operations
|
|
- Systemd hardening and process isolation
|
|
- Minimal attack surface (internal network only)
|
|
|
|
---
|
|
|
|
## Threat Model
|
|
|
|
### Threat Actor Profile
|
|
|
|
| Attribute | Description |
|
|
|-----------|-------------|
|
|
| **Origin** | Internal network only |
|
|
| **Skill Level** | Moderate to High |
|
|
| **Resources** | Limited (not nation-state) |
|
|
| **Motivation** | Unauthorized system access, privilege escalation |
|
|
| **Access** | Must bypass mTLS + IP whitelist |
|
|
|
|
### STRIDE Threat Analysis
|
|
|
|
| Threat Category | Potential Threat | Mitigation | Status |
|
|
|-----------------|------------------|------------|--------|
|
|
| **Spoofing** | Attacker impersonates valid client | mTLS certificate validation, unique certs per client | ✅ Mitigated |
|
|
| **Spoofing** | Attacker uses expired/revoked cert | Certificate expiry validation (1-year max) | ✅ Mitigated |
|
|
| **Tampering** | API requests modified in transit | TLS 1.3 encryption | ✅ Mitigated |
|
|
| **Tampering** | Config files modified unauthorized | File permissions (600/644), config validation before reload | ✅ Mitigated |
|
|
| **Repudiation** | Client denies making request | Audit logging with request_id, client cert ID | ✅ Mitigated |
|
|
| **Repudiation** | Server denies response | Comprehensive audit trail (systemd journal) | ✅ Mitigated |
|
|
| **Information Disclosure** | Package/data info leaked to unauthorized | Silent drop for non-mTLS, IP whitelist | ✅ Mitigated |
|
|
| **Information Disclosure** | Error messages leak system info | Detailed errors only for authenticated clients | ✅ Mitigated |
|
|
| **Denial of Service** | Resource exhaustion via many requests | Internal network only, IP whitelist limits exposure | ⚠️ Partial |
|
|
| **Denial of Service** | Job queue flooding | Configurable concurrent job limit (default: 5) | ✅ Mitigated |
|
|
| **Denial of Service** | Long-running job starvation | 30-minute job timeout enforcement | ✅ Mitigated |
|
|
| **Elevation of Privilege** | Unauthorized package installation | Root required, but mTLS + IP whitelist required | ✅ Mitigated |
|
|
| **Elevation of Privilege** | Subprocess escape | SystemCallFilter, ProtectSystem=strict | ✅ Mitigated |
|
|
|
|
### Attack Vectors & Mitigations
|
|
|
|
| Attack Vector | Likelihood | Impact | Mitigation |
|
|
|---------------|------------|--------|------------|
|
|
| Network interception | Low | Critical | TLS 1.3 only, mTLS required |
|
|
| Certificate theft | Medium | Critical | Cert permissions (600), internal CA only |
|
|
| IP spoofing | Low | High | IP whitelist + mTLS (both required) |
|
|
| Config file tampering | Medium | High | File permissions, validation before reload |
|
|
| Package manager injection | Low | Critical | Pluggable backend with input validation |
|
|
| Job manipulation | Low | High | Job storage isolation, exclusive rollback mode |
|
|
| Log tampering | Medium | High | systemd journal (immutable), optional remote syslog |
|
|
|
|
---
|
|
|
|
## Authentication & Authorization
|
|
|
|
### Authentication Requirements
|
|
|
|
- **Method:** mTLS certificate-based authentication
|
|
- **Certificate Type:** Unique client certificate per client (1-year validity)
|
|
- **CA:** Internal self-hosted Certificate Authority
|
|
- **TLS Version:** TLS 1.3 only
|
|
- **Multi-factor:** Certificate + IP whitelist (dual requirement)
|
|
- **Session Management:** Stateless (no sessions)
|
|
|
|
### Authorization Model
|
|
|
|
- **Model:** Binary authorization (all-or-nothing)
|
|
- **Permission Levels:** Single level (full access if authenticated)
|
|
- **Requirements:**
|
|
- Valid mTLS certificate (not expired, signed by internal CA)
|
|
- Source IP in whitelist (YAML config, instant apply)
|
|
- **No RBAC:** All authenticated clients have full API access
|
|
|
|
---
|
|
|
|
## Data Security
|
|
|
|
### Encryption at Rest
|
|
|
|
- **Certificates:** File permissions 600 for private keys
|
|
- **Job Storage:** `/var/lib/linux_patch_api/jobs/` (cleared on restart)
|
|
- **Config Files:** `/etc/linux_patch_api/` (644 for config, 600 for keys)
|
|
- **Audit Logs:** systemd journal (immutable by default)
|
|
|
|
### Encryption in Transit
|
|
|
|
- **Protocol:** TLS 1.3 only
|
|
- **Port:** 12443
|
|
- **Cipher Suites:** TLS 1.3 default (no legacy cipher support)
|
|
- **Certificate Validation:** Mutual TLS (server + client cert required)
|
|
|
|
### Key Management
|
|
|
|
- **CA Private Key:** Stored securely on CA host only
|
|
- **Server Certificates:** `/etc/linux_patch_api/certs/server.key` (600)
|
|
- **Client Certificates:** Distributed manually to authorized clients
|
|
- **Rotation:** 1-year certificate expiry, manual renewal process
|
|
- **Revocation:** Not implemented (rely on expiry + physical cert retrieval)
|
|
|
|
---
|
|
|
|
## API Security
|
|
|
|
### Input Validation
|
|
|
|
- **Package Names:** Alphanumeric + standard package chars only
|
|
- **Versions:** Semantic versioning validation
|
|
- **IP Addresses:** IPv4 + CIDR validation for whitelist
|
|
- **JSON Schema:** Strict schema validation for all request bodies
|
|
- **Path Traversal:** Blocked (no `..` in paths)
|
|
|
|
### Rate Limiting
|
|
|
|
- **Not Required:** Internal network only with strict IP whitelist
|
|
- **Job Concurrency:** Configurable limit (default: 5 concurrent jobs)
|
|
- **Job Timeout:** 30-minute maximum per job
|
|
|
|
### CORS Policy
|
|
|
|
- **Not Applicable:** API is not browser-accessible
|
|
- **Origin:** mTLS clients only (no browser CORS concerns)
|
|
|
|
---
|
|
|
|
## Audit & Logging
|
|
|
|
### Security Events to Log
|
|
|
|
- All API requests (endpoint, method, timestamp, client cert ID, source IP)
|
|
- Authentication events (success/failure, cert validation result)
|
|
- Authorization events (IP whitelist match/failure)
|
|
- Package operations (package name, version, action, result)
|
|
- Configuration changes (config reload, whitelist updates)
|
|
- Job lifecycle events (create, start, complete, fail, timeout, rollback)
|
|
- Service events (start, stop, restart, config validation failures)
|
|
|
|
### Log Protection
|
|
|
|
- **Primary Storage:** systemd journal (immutable, access-controlled)
|
|
- **Secondary Storage:** Optional remote syslog
|
|
- **Fallback:** Local file `/var/log/linux_patch_api/audit.log` (640)
|
|
- **Retention:** 30 days with daily rotation and compression
|
|
- **Access:** Root only, audit group read access
|
|
- **Integrity:** systemd journal provides tamper evidence
|
|
|
|
---
|
|
|
|
## Compliance Requirements
|
|
|
|
- **Internal Standards:** Follows organizational security policies
|
|
- **No External Compliance:** Not designed for PCI-DSS, HIPAA, SOC2 (can be extended)
|
|
- **Audit Trail:** Comprehensive logging supports internal audit requirements
|
|
- **Access Control:** mTLS + IP whitelist provides strong access control
|
|
|
|
---
|
|
|
|
## Security Testing
|
|
|
|
### Penetration Testing
|
|
|
|
- **Schedule:** Required before production deployment
|
|
- **Scope:**
|
|
- mTLS authentication bypass attempts
|
|
- IP whitelist enforcement testing
|
|
- API endpoint fuzzing
|
|
- Certificate validation testing
|
|
- Config file tampering attempts
|
|
- Privilege escalation attempts
|
|
- **Tester:** Internal security team or external contractor
|
|
- **Frequency:** Annual or after major changes
|
|
|
|
### Vulnerability Management
|
|
|
|
- **Dependency Scanning:** Rust crate security advisories monitored
|
|
- **System Patches:** Host system patched via API itself (dogfooding)
|
|
- **Certificate Updates:** Annual renewal process
|
|
- **Config Audits:** Quarterly review of whitelist and security settings
|
|
- **Incident Response:** Log analysis for security event investigation
|
|
|
|
---
|
|
|
|
*Following kiro spec-driven development standards*
|
|
*Following kiro spec-driven development standards*
|