DeepTx: Real-Time Transaction Risk Analysis via Multi-Modal Features and LLM Reasoning

TL;DR

DeepTx is a real-time system that leverages multi-modal features and large language models to detect Web3 phishing transactions with high accuracy, providing explainable threat assessments before user confirmation.

Contribution

It introduces a novel real-time transaction analysis framework combining multi-modal feature extraction with LLM reasoning and a self-reflective consensus mechanism.

Findings

High precision and recall in phishing detection

Effective simulation and feature extraction from transactions

Robust and explainable decision-making process

Abstract

Phishing attacks in Web3 ecosystems are increasingly sophisticated, exploiting deceptive contract logic, malicious frontend scripts, and token approval patterns. We present DeepTx, a real-time transaction analysis system that detects such threats before user confirmation. DeepTx simulates pending transactions, extracts behavior, context, and UI features, and uses multiple large language models (LLMs) to reason about transaction intent. A consensus mechanism with self-reflection ensures robust and explainable decisions. Evaluated on our phishing dataset, DeepTx achieves high precision and recall (demo video: https://youtu.be/4OfK9KCEXUM).