PhishBENCH Dataset

This dataset was used for the purposes of our work titled: "PhishBENCH: A benchmark for legal tasksrelated to phishing attacks". It is a comprehensive legal-phishing benchmark based on diverse sources including legal documents, phishing scenarios, the GDPR and the Greek Penal Code. Synthetic queries and answers were constructed based on those documents via OpenAI's API with structured prompts, which were then evaluated and refined by legal experts. The final dataset comprises: True/False Questions (True_False_Questions.csv) Phishing Attack cases (phishing_scenarios.csv) Penal Code examples (articles.csv) Legal Cases (cases.csv) Guilty or Innocent Examples (cases_guilty_or_not.csv) Open-Question-Answering Examples (generative_queries.csv) The prompts used for the generation of this dataset are showcased in the Prompts for Data Generation.pdf file