# 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*