Dark Version of ChatGPT: FraudGPT Used for Cyber Attacks

Dark Version of ChatGPT: FraudGPT Used for Cyber Attacks

Recently, cybersecurity firm SlashNext discovered that hackers are exploiting AI chatbots to conduct phishing and malware attacks. A hacker known as CanadianKingpin12 is promoting a specialized AI chatbot called FraudGPT across various hacker forums.

Hackers Train Dark Software: FraudGPT

CanadianKingpin12 heavily markets FraudGPT on the dark web for the purpose of conducting online scams and hacking attacks. Unlike other AI chatbots, FraudGPT is primarily trained using various hacker datasets available on the dark web, along with large language models like ChatGPT and Google Bard (Large Language Models, LLM). Its main functionalities include launching sophisticated phishing attacks to steal credit card and password information, executing advanced social engineering attacks, exploiting cybersecurity vulnerabilities, creating and distributing malicious software, and employing 0-day vulnerabilities to attack systems.

Widespread Illegal Trading of ChatGPT on the Dark Web

Security research company Group-IB revealed that over 100,000 compromised ChatGPT accounts are being sold on the dark web market. Devices owned by these account holders have been infected with spyware, resulting in leaked account information. The number of compromised ChatGPT accounts has surged from 74 in June last year to 26,802 in May 2023. The majority of affected devices are in the Asia-Pacific region, followed by the Middle East and Africa, Europe, and Latin America. The countries with the most victims are India, Pakistan, and Brazil. As ChatGPT automatically saves user queries and AI responses, compromised accounts may expose users’ confidential or sensitive information, making them vulnerable to targeted attacks against businesses or employees. Group-IB’s latest research shows that ChatGPT accounts have become a hot target for hackers. The simplicity, practicality, and effectiveness of spyware make it one of hackers’ favorite tools. Spyware typically infects devices through phishing emails or other means to maximize the number of infected devices. Researchers advise users to be cautious about unknown emails, attachments, or links to avoid becoming victims.

Challenges to Information Security

The emergence of FraudGPT poses significant challenges to existing cybersecurity defenses. Its ability to automate large-scale phishing attacks, social engineering, and exploitation of various vulnerabilities can pose major threats to online security once widely deployed. Hence, we need to be vigilant about this new threat and seek effective defense measures. This requires not only a deeper understanding of AI technology by cybersecurity researchers but also collective efforts from policymakers, tech companies, and ordinary users to confront this challenge together.