Skip to content

KoroS11/NSAXxB_Backend

Repository files navigation

NSA-X: Neuro-Symbolic Autonomous Security Analyst (Experimental Phase)

** DISCLAIMER**: This is a research prototype with a working frontend UI and backend API. However, ML/Neural Network models are NOT implemented - the system uses symbolic rule-based processing only. Real-time ML anomaly detection is planned for future development.


Quick Start

Prerequisites

One-Command Setup

Windows (PowerShell):

git clone https://github.com/KoroS11/insight-navigator.git
cd insight-navigator
.\setup.ps1

Linux/macOS:

git clone https://github.com/KoroS11/insight-navigator.git
cd insight-navigator
chmod +x setup.sh && ./setup.sh

Running the Application

Terminal 1 - Backend:

cd backend
# Windows: .\venv\Scripts\Activate.ps1
# Linux/macOS: source venv/bin/activate
uvicorn app.main:app --reload

Terminal 2 - Frontend:

npm run dev

Open http://localhost:8080 and login with admin / admin123


1. Project Overview

NSA-X is a research-grade prototype for a neuro-symbolic security analysis system designed with the following principles:

  • Neuro-Symbolic Architecture: Combines neural network-based anomaly detection with symbolic rule-based reasoning for more robust and interpretable threat analysis
  • Human-in-the-Loop by Design: Security analysts remain the final decision authority; the system supports rather than replaces human judgment
  • Explainability-First: Every detection, recommendation, and action must be traceable and justifiable
  • Governance-Aware: Built-in audit trails, autonomy boundaries, and compliance controls

What This Prototype Demonstrates

This prototype visualizes the intended user experience and workflow for security analysts interacting with a neuro-symbolic detection system. It serves as:

  • A concept demonstration for stakeholders and mentors
  • A design reference for future implementation
  • A research artifact documenting the intended system behavior

2. What Is Implemented (Current State)

The following components exist today:

Component Status Description
Frontend UI ✅ Complete React + TypeScript + Tailwind CSS + shadcn/ui
Backend API ✅ Complete FastAPI + SQLAlchemy Async + JWT Auth
Database ✅ Complete SQLite (dev) / PostgreSQL (prod)
Event Pipeline ✅ Complete 7-layer processing pipeline (rule-based, no ML)
Symbolic Reasoning ✅ Complete Rule-based threat detection
Explainability Engine ✅ Complete Decision tree + counterfactuals
Audit System ✅ Complete Immutable audit trail
Authentication ✅ Complete JWT + OAuth2 password flow
Test Coverage ✅ 87% 192 tests passing

Technical Stack

Layer Technology
Frontend React 18, TypeScript, TanStack Query, Tailwind CSS, shadcn/ui
Backend FastAPI, Python 3.11+, Pydantic v2
Database SQLAlchemy 2.0 Async, SQLite/PostgreSQL
Auth python-jose (JWT), passlib (bcrypt)
Testing pytest, pytest-asyncio, 87% coverage

Important Clarifications

  • Frontend and Backend are fully connected — Real API calls, not mocked
  • Data is persisted — Events, alerts, decisions stored in database
  • Authentication works — JWT-based login with admin/admin123
  • ML/Neural Networks NOT implemented — Anomaly detection is rule-based only

3. What Is NOT Implemented

This section is critical for setting accurate expectations.

Component Status Notes
ML Detection Models ❌ Not implemented No neural network anomaly detection
Behavioral Baseline ❌ Not implemented No learned user/entity behavior profiles
Sequence Analysis ❌ Not implemented No LSTM/Transformer for event sequences
Real Anomaly Scoring ❌ Not implemented Scores are rule-based, not ML-generated
Data Normalization ❌ Not implemented No CEF/STIX/OCSF parsing
SIEM Integration ❌ Not implemented No Splunk/Sentinel/Elastic connectors
SOAR Integration ❌ Not implemented No Phantom/XSOAR/Tines integration
Ticketing Integration ❌ Not implemented No Jira/ServiceNow connectivity
Alert Correlation ❌ Not implemented No event grouping or chaining
Threat Intelligence ❌ Not implemented No IOC enrichment

What the UI Shows vs. What Actually Happens

UI Element Visual State Actual Behavior
Confidence Score Shows percentage Calculated from rules, not ML
Event Timeline Shows events Real events from database
Explanation Tree Shows decision path Generated from symbolic reasoning
Audit Trail Shows log entries Persisted to database
Analyst Notes Accepts input Saved to database

Additional Implementations (Planned / To Be Done)

The following components are designed and scoped but pending full implementation or integration.

Component Planned Scope
Neural Detection Layer Behavioral baseline, sequence analysis, anomaly scoring with ML models
Frontend UI Prototype Full SOC-style React UI with TypeScript & Tailwind CSS
Design System Dark theme, SOC-style layout, consistent component library
Dashboard Layout Main navigation, sidebar, real-time status indicators
Architecture Visualization Conceptual and interactive neuro-symbolic pipeline diagrams
Mock Security Events Realistic synthetic event generation with proper identifiers
Analyst Decision Interface Confidence indicators, action matrix, analyst notes panel
Explainability UI Event summaries, explanation trees, counterfactual visualization
Governance Interface Audit trail viewer, autonomy boundary controls
Playbook Reference Static and dynamic playbook listing for analyst guidance

4. Intended System Architecture (Conceptual)

The following describes the target architecture, not the current implementation.

┌─────────────────────────────────────────────────────────────────┐
│                      DATA INGESTION LAYER                        │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐        │
│  │   SIEM   │  │   EDR    │  │ Network  │  │   IAM    │        │
│  │  Events  │  │  Alerts  │  │  Flows   │  │   Logs   │        │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘  └────┬─────┘        │
│       └──────────────┴──────────────┴──────────────┘            │
│                              │                                   │
│                    ┌─────────▼─────────┐                        │
│                    │   Normalization   │                        │
│                    │   (CEF → OCSF)    │                        │
│                    └─────────┬─────────┘                        │
└──────────────────────────────┼──────────────────────────────────┘
                               │
┌──────────────────────────────▼──────────────────────────────────┐
│                     NEURAL DETECTION LAYER                       │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐ │
│  │ Behavioral      │  │ Sequence        │  │ Anomaly         │ │
│  │ Baseline        │  │ Analysis        │  │ Scoring         │ │
│  └────────┬────────┘  └────────┬────────┘  └────────┬────────┘ │
│           └─────────────────────┴─────────────────────┘         │
│                              │                                   │
│                    ┌─────────▼─────────┐                        │
│                    │ Detection Output  │                        │
│                    │ + Confidence      │                        │
│                    └─────────┬─────────┘                        │
└──────────────────────────────┼──────────────────────────────────┘
                               │
┌──────────────────────────────▼──────────────────────────────────┐
│                   SYMBOLIC REASONING LAYER                       │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐ │
│  │ Rule Engine     │  │ Policy DSL      │  │ Conflict        │ │
│  │ (Rete/CLIPS)    │  │ Evaluation      │  │ Resolution      │ │
│  └────────┬────────┘  └────────┬────────┘  └────────┬────────┘ │
│           └─────────────────────┴─────────────────────┘         │
│                              │                                   │
│                    ┌─────────▼─────────┐                        │
│                    │ Reasoned Decision │                        │
│                    │ + Rule Trace      │                        │
│                    └─────────┬─────────┘                        │
└──────────────────────────────┼──────────────────────────────────┘
                               │
┌──────────────────────────────▼──────────────────────────────────┐
│                    EXPLAINABILITY ENGINE                         │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐ │
│  │ Provenance      │  │ Evidence        │  │ Counterfactual  │ │
│  │ Tracking        │  │ Attribution     │  │ Generation      │ │
│  └────────┬────────┘  └────────┬────────┘  └────────┬────────┘ │
│           └─────────────────────┴─────────────────────┘         │
│                              │                                   │
│                    ┌─────────▼─────────┐                        │
│                    │ Human-Readable    │                        │
│                    │ Explanation       │                        │
│                    └─────────┬─────────┘                        │
└──────────────────────────────┼──────────────────────────────────┘
                               │
┌──────────────────────────────▼──────────────────────────────────┐
│                   ANALYST DECISION LAYER                         │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐ │
│  │ Decision        │  │ Escalation      │  │ Feedback        │ │
│  │ Interface       │  │ Workflow        │  │ Loop            │ │
│  └─────────────────┘  └─────────────────┘  └─────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
                               │
┌──────────────────────────────▼──────────────────────────────────┐
│                    GOVERNANCE & AUDIT                            │
│  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐ │
│  │ Audit Trail     │  │ Autonomy        │  │ Compliance      │ │
│  │ (Immutable)     │  │ Boundaries      │  │ Reporting       │ │
│  └─────────────────┘  └─────────────────┘  └─────────────────┘ │
└─────────────────────────────────────────────────────────────────┘

Layer Descriptions

Layer Purpose Key Components
Data Ingestion Collect and normalize security telemetry SIEM connectors, log parsers, OCSF normalization
Neural Detection Identify anomalies using ML models Behavioral baselines, sequence models, anomaly scorers
Symbolic Reasoning Apply rules and policies to detections Rule engine, policy DSL, conflict resolution
Explainability Generate human-understandable explanations Provenance graphs, evidence chains, counterfactuals
Analyst Decision Present findings for human review Decision UI, escalation workflows, feedback capture
Governance Ensure accountability and compliance Immutable audit logs, autonomy controls, reports

5. MVP TODO List

What Needs to Be Built to Reach MVP

This is the prioritized work required to transform the prototype into a functional minimum viable product.


5.1 Backend Infrastructure

Task Why Needed Complexity
Design database schema Persist events, decisions, audit logs Medium
Implement event ingestion API Receive security events from sources Medium
Build normalization service Convert diverse formats to OCSF High
Create alert generation logic Transform anomalies into actionable alerts Medium
Set up authentication system Secure access, enable RBAC Medium
Implement API layer Connect frontend to backend services Medium
Add logging and monitoring Operational visibility Low

5.2 Detection Capabilities

Task Why Needed Complexity
Implement baseline anomaly detection Core detection capability High
Build feature extraction pipeline Prepare data for ML models High
Create model training workflow Enable model updates High
Implement drift detection Detect when models degrade Medium
Add detection confidence scoring Quantify detection reliability Medium

5.3 Symbolic Reasoning

Task Why Needed Complexity
Select and integrate rule engine Execute symbolic rules (Rete, CLIPS, Drools) High
Design policy DSL Allow analysts to define rules High
Implement conflict resolution Handle contradictory rules Medium
Build rule versioning Track policy changes Low
Create rule testing framework Validate rules before deployment Medium

5.4 Explainability

Task Why Needed Complexity
Implement provenance tracking Trace data lineage High
Build evidence attribution Link conclusions to supporting data High
Create explanation templates Generate consistent explanations Medium
Implement counterfactual generation Show "what would change the outcome" High
Add confidence breakdown Explain score components Medium

5.5 Analyst Workflows

Task Why Needed Complexity
Persist analyst decisions Save verdicts and justifications Low
Implement escalation workflows Route to appropriate responders Medium
Build feedback capture Enable model improvement Medium
Add case management Group related alerts Medium
Create analyst performance metrics Measure decision quality Low

5.6 Governance

Task Why Needed Complexity
Implement immutable audit log Regulatory compliance Medium
Build autonomy boundary enforcement Prevent unauthorized automation Medium
Create compliance reporting Generate audit reports Low
Add role-based access control Restrict sensitive operations Medium

5.7 Integrations

Task Why Needed Complexity
SIEM connector (Splunk/Sentinel) Ingest security events High
EDR connector (CrowdStrike/Defender) Endpoint telemetry High
SOAR integration Automate response actions High
Ticketing integration (Jira/ServiceNow) Track incidents Medium
Threat intel feeds Enrich with IOCs Medium

6. Roadmap

A realistic phased approach to reach production readiness.

Phase 0: Frontend Prototype ✅ COMPLETE

  • UI/UX design and component library
  • Mock data and workflow visualization
  • Architecture documentation
  • Stakeholder demonstration

Phase 1: Core Backend + Basic Detection

  • Database schema and API layer
  • Event ingestion and normalization
  • Simple anomaly detection (threshold-based)
  • Basic persistence (decisions, audit logs)
  • Authentication and basic RBAC

Exit Criteria: System can ingest events, detect simple anomalies, and persist analyst decisions.

Phase 2: Explainability + Analyst Workflows

  • Rule engine integration
  • Provenance tracking
  • Evidence attribution
  • Escalation workflows
  • Feedback loop

Exit Criteria: Detections include explanations; analysts can review, decide, and provide feedback.

Phase 3: Pilot-Ready System

  • ML-based detection models
  • Counterfactual generation
  • SIEM/EDR integration
  • Compliance reporting
  • Performance optimization

Exit Criteria: System can be deployed in a controlled environment for analyst evaluation.

Phase 4: Production Hardening (Future)

  • High availability architecture
  • Security audit and penetration testing
  • Scalability testing
  • Full RBAC implementation
  • Enterprise integrations

7. Who This Is For

Primary Audience

Audience Use Case
SOC Analysts Evaluate the proposed workflow and provide feedback
Security Architects Assess the system design and integration points
Researchers Study neuro-symbolic approaches to security
Mentors/Judges Evaluate the project's technical merit and feasibility

Not Intended For

  • End users seeking a production tool — This is a research prototype
  • Organizations needing immediate deployment — Significant development required
  • Non-technical stakeholders — Documentation assumes security domain knowledge

8. Project Structure

├── public/
│   ├── favicon.ico
│   ├── placeholder.svg
│   └── robots.txt
├── src/
│   ├── components/
│   │   ├── architecture/
│   │   │   └── ArchitectureFlow.tsx      # System architecture diagram
│   │   ├── decisions/
│   │   │   ├── ActionMatrix.tsx          # Decision action recommendations
│   │   │   ├── AnalystNotes.tsx          # Analyst note-taking (not persisted)
│   │   │   └── ConfidenceIndicator.tsx   # AI confidence visualization
│   │   ├── explainability/
│   │   │   ├── CounterfactualPanel.tsx   # "What-if" scenarios (static)
│   │   │   ├── EventSummary.tsx          # Event details summary
│   │   │   └── ExplanationTree.tsx       # Decision tree visualization
│   │   ├── governance/
│   │   │   ├── AuditTrail.tsx            # Audit logs (mock data)
│   │   │   └── AutonomyBoundary.tsx      # AI autonomy controls
│   │   ├── layout/
│   │   │   ├── AppLayout.tsx             # Main application layout
│   │   │   ├── AppSidebar.tsx            # Navigation sidebar
│   │   │   └── StatusBar.tsx             # System status indicator
│   │   ├── shared/
│   │   │   ├── MetricCard.tsx            # Reusable metric display
│   │   │   └── StatusBadge.tsx           # Status indicator badge
│   │   ├── ui/                           # shadcn/ui components
│   │   └── NavLink.tsx                   # Navigation link component
│   ├── hooks/
│   │   ├── use-mobile.tsx                # Mobile detection hook
│   │   └── use-toast.ts                  # Toast notification hook
│   ├── integrations/
│   │   └── supabase/
│   │       ├── client.ts                 # Supabase client (not used yet)
│   │       └── types.ts                  # Database types (empty)
│   ├── lib/
│   │   └── utils.ts                      # Utility functions
│   ├── pages/
│   │   ├── Architecture.tsx              # System architecture page
│   │   ├── Decisions.tsx                 # Analyst decision interface
│   │   ├── Explainability.tsx            # AI explanation dashboard
│   │   ├── Governance.tsx                # Governance & compliance
│   │   ├── Index.tsx                     # Main dashboard
│   │   └── NotFound.tsx                  # 404 error page
│   ├── App.tsx                           # Root application component
│   ├── App.css                           # Global styles
│   ├── index.css                         # Tailwind CSS configuration
│   └── main.tsx                          # Application entry point
├── supabase/
│   └── config.toml                       # Supabase configuration
├── .env                                  # Environment variables
├── tailwind.config.ts                    # Tailwind configuration
├── vite.config.ts                        # Vite build configuration
└── package.json                          # Dependencies

9. Local Development

Prerequisites

  • Node.js 18+
  • npm or bun

Installation

# Clone the repository
git clone https://github.com/KoroS11/insight-navigator.git

# Navigate to project directory
cd insight-navigator

# Install dependencies
npm install

# Start development server
npm run dev

Available Scripts

Command Description
npm run dev Start development server
npm run build Build for production
npm run preview Preview production build
npm run lint Run ESLint

10. Changelog

All notable changes to this project are documented here.

[2026-01-12]

Documentation

  • Complete README rewrite with honest assessment of current state
  • Added detailed MVP TODO list with complexity ratings
  • Documented intended vs. actual architecture
  • Created phased roadmap

Added

  • Enabled Lovable Cloud backend integration (not yet utilized)

Fixed

  • Fixed TypeScript error in Decisions.tsx (removed unused recommendedAction prop)

[Initial Commit]

Added

  • Frontend UI prototype
  • Dashboard, Decisions, Explainability, Governance, Architecture pages
  • Component library with shadcn/ui
  • Mock data for demonstration
  • Dark theme design system

11. License

License not yet specified. All rights reserved.


12. Contributing

  1. Create a feature branch from main
  2. Make your changes
  3. Submit a pull request for review

This is a research prototype developed to explore neuro-symbolic approaches to security operations. It is not intended for production use without significant additional development.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors