GK
AboutSkillsProjectsCertificatesContact
00:00:00
AboutSkillsProjectsCertificatesContact
Crafted with pixels & logic© 2026 Gautham Krishna. All rights reserved.
Sticky-Net
Back to Projects

Sticky-Net

AI Honeypot System

pythonPythonfastapiFastAPIGemini AIdockerDockerCloud RunNLPRegex

An AI-powered honeypot system that autonomously detects scam messages, engages scammers through multi-turn conversations, and extracts actionable intelligence for law enforcement. It uses a hybrid regex + AI pipeline to classify threats in under 150ms and adaptively engages scammers to extract bank accounts, UPI IDs, and beneficiary names.

Unique Features

4-stage pipeline: Regex pre-filter (~10ms) → AI Classifier (Gemini Flash, ~150ms) → Engagement Policy → Intelligence Extraction
Adaptive engagement modes: Cautious (10 turns) vs Aggressive (25 turns) with persona maintenance
Extracts 10+ intelligence types: bank accounts, UPI IDs, phone numbers, beneficiary names, IFSC codes, phishing URLs, crypto wallets
Shadow DOM isolation for injected UI components
Smart exit conditions with stale conversation detection and monotonic confidence scoring
Supports 50+ UPI provider and 50+ Indian bank name validation patterns

Tech Stack

FrontendReact 19, Tailwind CSS, shadcn/ui, Framer Motion
BackendFastAPI, Python 3.11+
AI/MLGemini Flash, Regex-based pre-filter, NLP
InfrastructureGoogle Cloud Run, Docker, Firestore, CORS middleware