Fraud-R1 : A Multi-Round Benchmark for Assessing the Robustness of LLM Against Augmented Fraud and Phishing Inducements

TL;DR

Fraud-R1 is a comprehensive benchmark with multi-round evaluation to test LLMs' robustness against various internet fraud and phishing tactics across different languages and interaction settings.

Contribution

Introduces Fraud-R1, a multi-round benchmark with diverse fraud cases and evaluation scenarios to assess LLMs' resistance to online fraud and phishing.

Findings

LLMs struggle more in role-play settings and with fake job postings.

Significant performance gap exists between Chinese and English LLMs.

Multi-round evaluation reveals vulnerabilities in LLM defenses against fraud.

Abstract

We introduce Fraud-R1, a benchmark designed to evaluate LLMs' ability to defend against internet fraud and phishing in dynamic, real-world scenarios. Fraud-R1 comprises 8,564 fraud cases sourced from phishing scams, fake job postings, social media, and news, categorized into 5 major fraud types. Unlike previous benchmarks, Fraud-R1 introduces a multi-round evaluation pipeline to assess LLMs' resistance to fraud at different stages, including credibility building, urgency creation, and emotional manipulation. Furthermore, we evaluate 15 LLMs under two settings: 1. Helpful-Assistant, where the LLM provides general decision-making assistance, and 2. Role-play, where the model assumes a specific persona, widely used in real-world agent-based interactions. Our evaluation reveals the significant challenges in defending against fraud and phishing inducement, especially in role-play settings and fake job postings. Additionally, we observe a substantial performance gap between Chinese and English, underscoring the need for improved multilingual fraud detection capabilities.