Security & Compliance

Document Version: v1.9.0
Last Updated: August 8, 2026
Audience: Enterprise teams, security reviewers, compliance officers

Overview

This document provides a detailed overview of DevDiff's security architecture and compliance posture for organizations evaluating DevDiff for use in regulated or security-sensitive environments.

DevDiff is designed around three foundational security principles:

Zero Attack Surface — No DevDiff cloud servers means there are no DevDiff-operated systems to breach, no databases to exfiltrate, and no central point of failure.
Defense in Depth — Multiple independent, layered security controls protect your code context at every processing step.
Complete Auditability — Every security mechanism is open source, publicly auditable, and verifiable against a pinned lockfile.

Security Architecture

Data Flow Security

Developer Machine
  │
  ├─ Git Diff Extracted Locally
  │
  ├─ Injection Guard v2 ──── Sanitizes all user-controlled content
  │
  ├─ Redaction Engine v2 ─── Strips credentials, secrets, tokens
  │
  ├─ Cloud Guard ──────────── Validates cloud provider is explicitly configured
  │
  ├─ [Local Model] OR [Cloud Provider API] ── Your choice, your control
  │
  └─ AccuracyGuard ────────── Validates output against raw diff

No step in this pipeline involves DevDiff-operated infrastructure.

Security Controls

Control	Implementation File	Status
Credential Redaction	`packages/core/src/ai/redaction-engine-v2.ts`	✅ Active
Prompt Injection Prevention	`packages/core/src/security/injection-guard-v2.ts`	✅ Active
Network Egress Firewall	`packages/core/src/security/network-guard.ts`	✅ Active
Cloud Provider Guard	`packages/core/src/ai/cloud-guard.ts`	✅ Active
Output Accuracy Validation	`packages/core/src/ai/accuracy-guard.ts`	✅ Active
Path Traversal Prevention	Strict boundary validation on all filesystem operations	✅ Active
Shell Metacharacter Sanitization	Applied to all user-controlled CLI arguments and inputs	✅ Active
Encrypted Local Audit Logs	AES-256-GCM with PBKDF2 key derivation for `.devdiff/audit/`	✅ Active
MCP Server Authorization	Token-based authentication on all stdio and HTTP MCP endpoints	✅ Active
Secure File Permissions	`.env` and credential files created with `600` permissions	✅ Active
Dependency Integrity	pnpm frozen lockfile (`pnpm-lock.yaml`) enforced in all CI pipelines	✅ Active
CodeQL Static Analysis	Automated code scanning on every push to `main`	✅ Active
Secret Scanning	CI pre-release scan via `scripts/security-scan.js`	✅ Active
Dry-Run Transparency	`devdiff generate --dry-run` previews all data before any AI call	✅ Active

Compliance Posture

DevDiff is designed to operate entirely within the developer's local environment, which substantially reduces GDPR compliance exposure:

GDPR Requirement	DevDiff Posture
Lawful basis for processing	Not applicable — DevDiff processes no personal data belonging to end users
Data minimization	No personal data collected or retained
Right of access / erasure	Not applicable — no data held by DevDiff
Data transfers	No transfer to DevDiff infrastructure occurs
Sub-processor risk	If cloud AI providers are used, the developer's own DPA with that provider governs
Cookie consent	No cookies used on the documentation site

NOTE

When cloud AI providers are configured, data is transmitted directly from the developer's machine to the provider under the developer's own API credentials. DevDiff does not intermediate this transmission. Developers should review the Data Processing Addendum (DPA) of their chosen AI provider.

SOC 2 Type II Considerations

SOC 2 Trust Service Criteria	DevDiff Posture
Availability	Fully offline-capable; no operational dependency on DevDiff infrastructure uptime
Confidentiality	Code context never leaves the local environment by default; Network Guard enforces this
Processing Integrity	AccuracyGuard validates all AI-generated outputs against the source diff
Security	See Security Controls table above
Privacy	See Privacy Policy

HIPAA

CAUTION

DevDiff is not designed or approved for processing Protected Health Information (PHI). If your repository contains patient data or healthcare records, you must:

Configure DevDiff to use a local-only AI provider (Ollama) to ensure zero network egress.
Verify that your .devdiff.config.js has no cloud AI providers configured.
Consult your organization's compliance team before using any AI tooling on PHI-adjacent repositories.

Air-Gapped & Offline Environments

DevDiff is fully operational in air-gapped or offline environments:

bash

# Install CLI on air-gapped machine from a pre-downloaded package tarball
npm install -g /path/to/@eldrex-cli-1.5.0.tgz

# Install VS Code Extension from VSIX file
# Available at: GitHub Releases → devdiff-vscode-1.5.0.vsix
# Install via: VS Code → Extensions → "Install from VSIX..."

All core functionality — diff analysis, changelog generation, persistent memory indexing, and Q&A — operates completely offline when a local AI provider (Ollama) is used.

For detailed deployment guidance, see the Air-Gapped Environments guide.

Vulnerability Disclosure

DevDiff follows a coordinated responsible disclosure model. Full details are documented in the Security Policy.

CAUTION

Do not open public GitHub issues for security vulnerabilities. Public disclosure before a patch is released puts all users at risk.

To report a vulnerability:

Navigate to GitHub Security Advisories
Click "Report a vulnerability"
Provide reproduction steps, affected versions, and impact assessment

Response Timeline:

Stage	Target SLA
Acknowledgment	Within 48 hours
Triage & Assessment	Within 5 business days
Critical Fix	Within 14 days
Standard Fix	Within 90 days
Public Disclosure	After fix is released

Scope:

In scope: @eldrex/core, @eldrex/cli, @eldrex/gateway, @eldrex/mcp, VS Code Extension
Out of scope: Third-party AI provider security, social engineering, physical access

Security Audit Trail

All security-relevant changes are tracked in the public git history and release notes:

Version	Date	Security Changes
v1.5.0	2026-08-07	Cloud Guard explicit opt-in enforcement, PersistentMemory audit trail, enhanced Injection Guard patterns
v1.0.6	2026-07-06	Windows argument parsing hardening, MCP port security improvements
v1.0.3	2026-06-28	Initial Redaction Engine v2, Network Guard firewall implementation

Contact

For security-related inquiries or compliance documentation requests:

Channel	Link
Private Security Advisory	GitHub Security Advisories
General Issues	GitHub Issues
Email	eldrexdelosreyesbula@gmail.com

Security & Compliance ​

Overview ​

Security Architecture ​

Data Flow Security ​

Security Controls ​

Compliance Posture ​

GDPR (General Data Protection Regulation) ​

SOC 2 Type II Considerations ​

HIPAA ​

Air-Gapped & Offline Environments ​

Vulnerability Disclosure ​

Security Audit Trail ​

Contact ​