AI Threat Detective

Gallery

About

AI Threat DetectiveNeural engine for real-time anomaly detection in DeFi/Web3 transactions.AI Threat Detective analyzes transaction windows and returns an anomaly score plus business-ready explanations (risk factors + recommendations) so teams can act before funds are drained.How it works (under the hood)Input: transactions (single tx or window) via REST or WebSocketFeature engineering: tabular window features + graph metrics (from→to)Model scoring: Transformer + GNN + Tabular modelsFusion: mean-based fusion across available modules (safe fallback if some artifacts are missing)Decision: is_anomaly based on a configurable thresholdXAI output: summary, risk_factors, recommendations for CFO/CISO/opsWhat’s includedFull source codeREST API + WebSocket streamingXAI layer for human-readable explanations (DeFi-focused templates)Fusion ML architecture: Transformer + GNN + Tabular + fallback-safe scoringDocker image + artifacts workflowOptional alerting via webhook when anomalies are detectedAPI keys + rate limiting (for multi-tenant / customer usage)White-label rights: rebrand and resell as your own serviceWhy buySave months of engineering: built as an integration-focused AI engine, not a research prototypeFaster deployment: deploy and integrate using Docker + provided endpoints and examplesPractical explainability: explanations are designed to be understandable by non-ML stakeholdersMulti-chain ready: chain_id support + adapter pattern (Ethereum mainnet streaming included as an example)Important note on model trainingCurrent training scripts generate synthetic MVP datasets. The architecture is designed for fine-tuning and re-training on your own labeled or customer datasets for production calibration.Tech stackPython, PyTorch, PyTorch Geometric, FastAPI, scikit-learn, DockerPrice$3,500 one-time license (white-label rights included)Demo video available upon request (includes live start + UI tests).