CYBIKHAN-AI: The Cognitive Firewall
Explainable Multi-Modal AI – Protecting Human Decisions, Not Just Devices.
CYBIKHAN-AI is an explainable multi-modal cognitive defense framework built directly into the user's browser. It combines four distinct AI threat-detection modules — Textual Guard (NLP), Visual Deception Engine (CV), Sonic Shield (Acoustic Forensics), and URL Lexical Scanner — into one unified orchestration engine. Unlike traditional firewalls that scan packets, CYBIKHAN-AI monitors what the user sees and hears in real-time, stopping human-centric attacks before they can manipulate.
91%
Human-Centric Attacks Blocked
300%+
AI-Generated Threat Detection
< 3 sec
Response Time
24/7 Auto-Guard
Protection Mode
Three-Tier Distributed Architecture
CYBIKHAN-AI operates on a highly optimized distributed architecture, keeping heavy AI processing off the user's local machine to ensure zero browser lag.
Tier 1: Client (Chrome Extension)
React.js, Vite, Tailwind CSS, IndexedDB (Dexie) for local data storage
Tier 2: Middleware Bridge
Node.js + Express for heavy payload routing and threat telemetry logging
Tier 3: AI Engine
FastAPI (Python) running DistilBERT, YOLOv8, and Librosa inference models
The 4 Pillars of CYBIKHAN-AI
Textual Guard (NLP)
Powered by DistilBERT, reads webpage text and emails bidirectionally to detect psychological manipulation, urgency bias, and authority fraud.
Visual Deception Engine
Utilizes YOLOv8 and custom CNNs to detect brand impersonation, fake logos, and deceptive UI elements in real-time.
Sonic Shield (Acoustic Forensics)
Analyzes Mel-Spectrograms with Librosa to detect synthetic, non-biological voice resonance and deepfake audio.
URL Lexical Scanner
ML classifiers analyze domain entropy, typosquatting, and structural anomalies of webpage URLs in real-time.
Explainable AI (XAI)
Uses LIME and SHAP to interpret complex AI decisions into human-readable language with visual heatmaps.
Threat Fusion Engine
Cross-verifies multiple modalities against each other to stop multi-vector attacks with zero false positives.
Core Technologies
Built with React.js and Tailwind CSS, the extension handles UI, data capture, and local storage using IndexedDB (Dexie).
Key Features
- React.js with Vite for fast development
- Tailwind CSS for responsive UI
- IndexedDB (Dexie) for storing 50MB+ base64 telemetry data
- Offscreen Document Pipeline for continuous audio monitoring
Specifications
Framework
React.js + Vite
Storage
IndexedDB (Dexie)
Intelligent Cognitive Defense Engine
CYBIKHAN-AI combines four distinct AI threat-detection modules under one orchestration engine to stop multi-vector attacks before the user can be manipulated.
Textual Guard (DistilBERT)
Reads webpage text and emails bidirectionally to detect urgency bias, authority fraud, and psychological manipulation.
Visual Deception Engine (YOLOv8 + CNNs)
Analyzes browser screenshots in real-time to detect brand impersonation, fake logos, and deceptive UI elements.
Sonic Shield (Librosa + STT)
Dual-layer acoustic forensics — analyzes Mel-Spectrograms for deepfake detection and transcribes audio for credential-harvesting intent.
URL Lexical Scanner
Machine Learning classifiers analyze domain entropy, typosquatting, and structural anomalies of webpage URLs in real-time.
Explainable AI (LIME + SHAP)
Interprets complex matrix decisions into human-readable language with visual heatmaps on deceptive elements.
Threat Fusion Engine
Cross-verifies modalities against each other (e.g., if voice is synthetic AND OTP request is detected, trigger alarm).
Technical Specifications
Applications & Target Audience
The CYBIKHAN-AI Advantage
Novel Technology & Features
Offscreen Document Pipeline
Bypasses Chrome MV3 limits — keeps the defense active 24/7 by chunking audio into 10-second heartbeats.
Threat Fusion Engine
Cross-verifies vision, text, and audio to stop multi-vector attacks with zero false positives.
Explainable AI (LIME + SHAP)
Translates complex AI decisions into human-readable explanations with visual heatmaps.
Visual Reference
CYBIKHAN-AI Browser Extension Dashboard
Explainable AI – Visual Heatmap Detection
Multi-Modal Threat Detection Overview