Skip to content

ROADMAP.md

Latest commit

 

History

History
622 lines (461 loc) · 19.1 KB

ROADMAP.md

File metadata and controls

622 lines (461 loc) · 19.1 KB

AIM Development Roadmap

Last Updated: December 24, 2025

This document tracks future enhancements and features that are deferred from the current development cycle.

✅ Recently Completed

Trust Score Policy Enforcement (December 2025)

  • Automatic enforcement when trust scores fall below thresholds
  • Score < 70%: Warning alert created (agent continues to operate)
  • Score < 50%: Critical alert created AND agent automatically suspended
  • Deduplication prevents duplicate alerts within 1 hour
  • Integrated into all trust score update paths

Java SDK (December 2025)

  • Full Java SDK with Maven/Gradle support
  • AspectJ @SecureAction annotations for declarative security
  • Spring Boot integration
  • OkHttp HTTP client with BouncyCastle Ed25519 cryptography
  • Tags and metadata support
  • Complete feature parity with Python SDK

Agent Tags & Metadata (December 2025)

  • Tags for categorization (auto-created if they don't exist)
  • Custom metadata as key-value pairs
  • Both Python and Java SDK support

Supply Chain Analytics Enhancements (December 2025)

  • ABOM (Agent Bill of Materials) tab
  • Capability drift alerts with pagination
  • Enhanced threat detection modal with tabs and smart recommendations

🛡️ Security Policy Enforcement (Post-MVP)

Current Status (MVP)

✅ Capability Violation Detection - FULLY IMPLEMENTED

  • Actively enforced on all agent API calls
  • Creates security alerts for violations
  • Blocks unauthorized actions in real-time
  • Prevents EchoLeak-style attacks (CVE-2025-32711)

Planned Security Policy Enforcers

Trust Score Policy Enforcement ✅ IMPLEMENTED

Priority: High Status: Implemented (December 2025)

Trust scores are now actively enforced with automatic policy evaluation:

Enforcement Rules:

  • Score < 70%: Warning alert created (agent continues to operate)
  • Score < 50%: Critical alert created AND agent automatically suspended

Implementation:

  • Added EvaluateTrustScoreOnUpdate() method to SecurityPolicyService
  • Called after every trust score update (recalculation, manual update, violation impact)
  • Creates alerts with deduplication (prevents duplicate alerts within 1 hour)
  • Automatically suspends agents with critical trust scores
  • Integrated into RecalculateTrustScore(), UpdateTrustScore(), and capability violation handling

Files Modified:

  • apps/backend/internal/application/security_policy_service.go
  • apps/backend/internal/application/agent_service.go
  • apps/backend/cmd/server/main.go

Failed Authentication Monitoring ✅

Priority: High Status: ✅ IMPLEMENTED

What Was Implemented:

  • ✅ auth_failures table to track failed login attempts (email, IP, user agent, metadata)
  • ✅ auth_lockouts table to track account lockouts with automatic unlock scheduling
  • ✅ AuthFailureRepository with full CRUD + lockout management
  • ✅ LoginWithPasswordExtended() in AuthService with lockout checking and failure recording
  • ✅ EvaluateAuthFailures() in SecurityPolicyService for alert creation
  • ✅ Automatic account lockout after 5 failed attempts within 15 minutes
  • ✅ 30-minute lockout duration (auto-unlock)
  • ✅ Warning alerts after 3 failures, high-severity alerts on lockout
  • ✅ Admin unlock capability via UnlockAccount() method
  • ✅ Duplicate alert prevention (cooldown periods)

Policy Configuration (default):

  • max_attempts: 5 failed attempts before lockout
  • time_window: 15 minutes for counting attempts
  • lockout_duration: 30 minutes
  • alert_threshold: 3 failures before warning alert

Key Files:

  • apps/backend/internal/domain/auth_failure.go
  • apps/backend/internal/infrastructure/repository/auth_failure_repository.go
  • apps/backend/migrations/062_add_auth_failures_monitoring.sql
  • apps/backend/internal/application/auth_service.go (updated)
  • apps/backend/internal/application/security_policy_service.go (updated)

Use Case: Prevent brute force attacks and credential stuffing

Unusual Activity Detection ✅

Priority: Medium Status: ✅ IMPLEMENTED

What Was Implemented:

  • ✅ BehaviorAnalysisService with per-agent baseline learning
  • ✅ AgentBehaviorBaseline domain model with velocity/capability/resource tracking
  • ✅ behavior_baselines and behavioral_anomalies tables (migration 055)
  • ✅ Statistical anomaly detection using sigma thresholds (2σ to 5σ)
  • ✅ EvaluateUnusualActivity() in SecurityPolicyService
  • ✅ Velocity spike detection (calls per hour vs baseline)
  • ✅ New capability/resource access detection
  • ✅ Risk-weighted activity scoring
  • ✅ Alert creation for detected anomalies
  • ✅ Configurable blocking via policy enforcement

Anomaly Types Detected:

  • velocity_spike - Sudden increase in activity rate
  • new_capability - First-time use of a capability
  • new_resource - Access to never-before-seen resource
  • pattern_break - Unusual action sequence
  • risk_spike - Sudden increase in risk-weighted score

Key Files:

  • apps/backend/internal/domain/behavior_baseline.go
  • apps/backend/internal/application/behavior_analysis_service.go
  • apps/backend/internal/infrastructure/repository/behavior_baseline_repository.go
  • apps/backend/migrations/055_add_agent_behavior_baselines.sql

Use Case: Detect compromised agents or malicious behavior

Data Exfiltration Detection ✅

Priority: Medium Status: ✅ IMPLEMENTED

What Was Implemented:

  • ✅ data_transfers table for individual transfer records
  • ✅ data_transfer_aggregates table for hourly rollups
  • ✅ DataTransferRepository with cumulative size tracking
  • ✅ Size-based detection in EvaluateDataExfiltration()
  • ✅ Pattern-based detection (fetch_external_url, bulk_export, etc.)
  • ✅ AlertDataExfiltration alert type
  • ✅ Configurable thresholds (default: 100MB per hour)
  • ✅ Destination IP/domain logging
  • ✅ Policy-based enforcement (block/alert/allow)

Configuration (via policy rules):

  • data_threshold_mb: Transfer size threshold (default: 100 MB)
  • time_window_mins: Time window for threshold (default: 60 minutes)
  • patterns: Suspicious action patterns to detect

Key Files:

  • apps/backend/internal/domain/data_transfer.go
  • apps/backend/internal/infrastructure/repository/data_transfer_repository.go
  • apps/backend/migrations/063_add_data_transfer_tracking.sql
  • apps/backend/internal/application/security_policy_service.go (updated)

Use Case: Prevent data breaches and insider threats

Deployment & Infrastructure

Docker Compose for Production

Priority: Medium Status: Deferred

Create a complete docker-compose.yml for single-command deployment:

  • PostgreSQL database with persistent volumes
  • Redis cache
  • Backend service
  • Frontend service
  • Auto-initialization on first run
  • Environment variable configuration
  • Health checks and restart policies

Use Case: Local production-like deployments and testing

GitHub Actions CI/CD Workflow

Priority: Medium Status: Deferred

Automate Docker image builds and deployments:

  • Build backend and frontend images on push to main
  • Push images to Azure Container Registry
  • Run tests before building
  • Automated deployment to Azure Container Apps
  • Multi-stage builds for optimization
  • Security scanning with Trivy

Use Case: Automated deployments on git push

One-Command Deployment Testing

Priority: Medium Status: Deferred

End-to-end testing of simplified deployment:

  • Test docker compose up deployment
  • Verify auto-initialization works
  • Validate all services start correctly
  • Test database migrations apply automatically
  • Verify admin user creation
  • Check default security policies seeded

Use Case: Ensuring deployment reliability

SDK Security & Stability Fixes

The following issues were identified during a comprehensive code review (December 2025) and are tracked for resolution.

Python SDK Issues

Memory Leak in LangChain Callback Handler ✅ FIXED

Priority: Medium Status: Fixed (December 2025) File: sdk/python/aim_sdk/integrations/langchain/callback.py

Issue: The _active_tools dictionary stores tool invocation data keyed by run_id. If on_tool_end or on_tool_error is never called for a tool (e.g., due to exception or unexpected flow), entries accumulate and are never cleaned up.

Fix: Added TTL-based cleanup mechanism with 1-hour expiry and max 1000 entries limit.

Silent Fallback to Unencrypted Storage ✅ FIXED

Priority: Medium Status: Fixed (December 2025) File: sdk/python/aim_sdk/oauth.py

Issue: When secure storage initialization fails (e.g., keyring unavailable), the code silently falls back to storing credentials in plaintext without warning the user.

Fix:

  • Added prominent UserWarning when falling back to plaintext storage
  • Added allow_plaintext_fallback parameter to OAuthTokenManager:
    • True (default): Falls back with warning
    • False: Raises RuntimeError requiring secure storage

JWT Token Decoding Without Signature Verification ✅ FIXED

Priority: High Status: Fixed (December 2025) File: sdk/python/aim_sdk/oauth.py

Issue: JWT tokens are decoded by splitting and base64-decoding the payload directly without cryptographic signature verification.

Fix:

  • Added PyJWT>=2.8.0 as a dependency
  • Created decode_jwt_claims() helper function using PyJWT for proper JWT parsing
  • PyJWT validates JWT structure (3 parts, valid base64, valid JSON payload)
  • Added issuer validation check (warns on unexpected issuers)
  • Replaced all manual base64 decoding with PyJWT

Note: Full signature verification is not performed locally because:

  • Server uses HS256 (symmetric key) - sharing the secret would defeat the purpose
  • Tokens are always verified server-side when used for API calls
  • Local parsing is only for housekeeping (reading expiry, token ID)

Java SDK Issues

HTTP Connection Pool Not Properly Closed ✅ FIXED

Priority: Medium Status: Fixed File: sdk/java/src/main/java/org/opena2a/aim/client/AIMClient.java

Issue: The close() method shuts down the executor service but does not evict connections from the OkHttpClient connection pool.

Fix: Fixed - evictAll() is called for both httpClient and authClient connection pools.

Race Condition in Key Rotation ✅ FIXED

Priority: High Status: Fixed (December 2025) File: sdk/java/src/main/java/org/opena2a/aim/security/SecureCredentialStorage.java

Issue: The rotateKey() method is not synchronized. Concurrent calls could corrupt credentials or the key file.

Fix: Added synchronized keyword to rotateKey() method to prevent concurrent execution.

Master Key Security Concerns ✅ PARTIALLY FIXED

Priority: High Status: Partially Fixed (December 2025) File: sdk/java/src/main/java/org/opena2a/aim/security/SecureCredentialStorage.java

Issue:

  1. Master encryption key stored in plain file (~/.aim/secure/.keystore)
  2. Key is never zeroed from memory after use
  3. No integrity check on the key file

Fixes Applied:

  • ✅ deriveKey() now zeros PBEKeySpec password and intermediate key bytes after use
  • ✅ Added clearMasterKey() method to securely zero master key from memory
  • ✅ Auto-registers JVM shutdown hook to clear key on exit
  • ✅ rotateKey() already zeros old key after rotation

Still TODO:

  • OS keychain integration (macOS Keychain, Windows Credential Manager)
  • HMAC integrity check for key file

No Retry Logic in Client Credentials OAuth Flow ✅ FIXED

Priority: Medium Status: Already Fixed File: sdk/java/src/main/java/org/opena2a/aim/client/AIMClient.java

Issue: The authenticateWithClientCredentials() method has no retry logic for transient network failures.

Fix: Fixed - method uses httpClient which has retry interceptor with exponential backoff.

Missing Agent Name Length Validation ✅ FIXED

Priority: Low Status: Fixed (December 2025) File: sdk/java/src/main/java/org/opena2a/aim/credentials/CredentialManager.java

Issue: Agent names are sanitized for filesystem safety but there's no maximum length validation, which could exceed filesystem path limits.

Fix: Added 200-character maximum length check with clear error message.

🔐 Security Enhancements

Advanced RBAC System

Priority: High Status: Planned

Implement fine-grained role-based access control:

  • Custom role definitions
  • Permission-based access control
  • Role inheritance
  • Organization-level and resource-level permissions
  • Audit trail for role changes

Use Case: Organizations with complex permission requirements

Multi-Factor Authentication (MFA)

Priority: High Status: Planned

Add MFA support for enhanced security:

  • TOTP (Time-based One-Time Password)
  • SMS-based verification
  • Backup codes
  • Recovery mechanisms
  • Enforced MFA for admin accounts

Use Case: Compliance requirements (SOC 2, HIPAA)

API Rate Limiting ✅ IMPLEMENTED

Priority: Medium Status: Implemented (December 2025)

Features Implemented:

  • ✅ Per-user rate limits (authenticated users)
  • ✅ Per-organization rate limits (via callback)
  • ✅ Configurable via environment variables:
    • RATE_LIMIT_DEFAULT_MAX (default: 100)
    • RATE_LIMIT_DEFAULT_WINDOW_SEC (default: 60)
    • RATE_LIMIT_STRICT_MAX (default: 10)
    • RATE_LIMIT_STRICT_WINDOW_SEC (default: 60)
  • ✅ Rate limit headers in responses:
    • X-RateLimit-Limit
    • X-RateLimit-Remaining
    • X-RateLimit-Reset
    • Retry-After
  • ✅ Redis-based distributed rate limiting
  • ✅ Graceful fallback to in-memory when Redis unavailable
  • ✅ Strict rate limiting for sensitive endpoints (login, registration)

Files:

  • apps/backend/internal/interfaces/http/middleware/redis_rate_limit.go
  • apps/backend/internal/infrastructure/cache/redis.go

Use Case: Prevent abuse and ensure fair usage

Features & Enhancements

Advanced Analytics Dashboard

Priority: Medium Status: Planned

Enhanced analytics and insights:

  • Trust score trends over time
  • Agent usage patterns
  • Security incident heatmaps
  • Compliance reporting
  • Exportable reports (PDF, CSV)

Use Case: Security teams and compliance auditors

Webhook Integration System ✅ IMPLEMENTED

Priority: Medium Status: Complete

Allow external systems to receive AIM events:

  • ✅ Configurable webhook endpoints with CRUD API
  • ✅ Event filtering (16 event types: alerts, agents, trust scores, security, API keys)
  • ✅ Retry logic with exponential backoff (configurable max retries, delays)
  • ✅ Webhook signature verification (HMAC-SHA256)
  • ✅ Event replay capabilities
  • ✅ Delivery tracking and statistics
  • ✅ Soft delete support
  • ✅ Integration with alert system (alert.created, alert.acknowledged webhooks)

Implementation Details:

  • Enhanced domain model: apps/backend/internal/domain/webhook.go
  • Service with retry logic: apps/backend/internal/application/webhook_service.go
  • Repository implementation: apps/backend/internal/infrastructure/repository/webhook_repository.go
  • Migration: apps/backend/migrations/064_enhance_webhooks_for_retry.sql

Use Case: Integration with SIEM, Slack, PagerDuty, etc.

CLI Tool for Automation

Priority: Low Status: Planned

Command-line tool for AIM operations:

  • Agent registration via CLI
  • API key generation
  • Bulk operations
  • Configuration management
  • Scripting support

Use Case: DevOps automation and CI/CD pipelines

GraphQL API

Priority: Low Status: Planned

Add GraphQL endpoint alongside REST API:

  • Flexible querying
  • Reduced over-fetching
  • Real-time subscriptions
  • Schema introspection
  • GraphQL Playground

Use Case: Frontend flexibility and efficiency

Testing & Quality

Integration Test Suite

Priority: High Status: Planned

Comprehensive integration tests:

  • API endpoint tests
  • Database integration tests
  • Authentication flow tests
  • Authorization tests
  • Error handling tests

Use Case: Regression prevention and quality assurance

Load Testing Framework

Priority: Medium Status: Planned

Performance testing infrastructure:

  • k6 load testing scripts
  • Stress testing scenarios
  • Performance benchmarks
  • Scalability testing
  • Results visualization

Use Case: Ensuring performance at scale

E2E Frontend Tests

Priority: Medium Status: Planned

End-to-end UI testing:

  • Playwright test suite
  • Critical user journey tests
  • Cross-browser testing
  • Visual regression testing
  • Automated screenshot comparisons

Use Case: Frontend quality assurance

Framework Integrations

TypeScript/Node.js SDK

Priority: High Status: Planned (Q2 2026)

TypeScript SDK for Node.js applications:

  • Full TypeScript support with type definitions
  • async/await API design
  • npm package distribution
  • Express/Fastify middleware integration
  • Feature parity with Python and Java SDKs

Use Case: Node.js applications, serverless functions, TypeScript projects

GitHub Copilot Integration

Priority: Medium Status: Planned (Q2 2026)

Integrate AIM with GitHub Copilot for VS Code:

  • Copilot agent capability detection
  • Trust-based code review workflows
  • Audit trail for AI-generated code
  • Integration with GitHub Copilot Business/Enterprise

Implementation:

  • Copilot API hooks for code suggestions
  • Security policy enforcement for generated code
  • Dashboard visibility for Copilot usage patterns

Use Case: Organizations using GitHub Copilot who need visibility and control over AI-generated code

Documentation

API Documentation Portal

Priority: High Status: Planned

Interactive API documentation:

  • Swagger/OpenAPI spec
  • Interactive API explorer
  • Code examples in multiple languages
  • Authentication guide
  • Rate limiting documentation

Use Case: Developer onboarding and API adoption

Video Tutorials

Priority: Low Status: Planned

Video guides for common tasks:

  • Getting started with AIM
  • Registering your first agent
  • Configuring security policies
  • Integrating with SSO
  • Troubleshooting common issues

Use Case: User education and onboarding

Deployment History

Completed Deployments

  • December 22, 2025: Java SDK and documentation update

    • Java SDK with Maven/Gradle support
    • Tags and metadata support in both SDKs
    • Supply Chain Analytics enhancements
    • Updated README with dual-language examples
    • Migrated to Next.js 16

  • October 20, 2025: Auto-initialization feature deployed

    • Complete database schema
    • Default seed data
    • Automatic admin user creation
    • Super admin protection
    • Users page fixes
    • Organization settings fixes

Notes

  • Items in this roadmap are not prioritized in any particular order within their priority level

  • Priorities may change based on user feedback and business needs

  • Completed items will be moved to the "Completed Deployments" section

  • New items can be added by creating a PR to update this file

  • Some features may be removed from roadmap later

Questions or Suggestions? Open an issue on GitHub: https://github.com/opena2a-org/agent-identity-management/issues
⚡