Faux ChatGPT, Claude API Packages Deliver JarkaStealer

Faux ChatGPT, Claude API Packages Deliver JarkaStealer

November 22, 2024 at 05:08PM

Two malicious Python packages falsely marketed as tools for ChatGPT and Claude contained an infostealer named “JarkaStealer.” Designed to lure developers, they masqueraded as legitimate APIs but ultimately compromised users’ data. Over 1,700 downloads occurred before the packages were removed following discovery by Kaspersky researchers.

Here are the key takeaways from the meeting notes regarding the recent security breach involving malicious Python packages:

1. **Malware Discovery**: Two Python packages, “gptplus” and “claudeai-eng,” were found to contain malware, specifically an infostealer known as “JarkaStealer,” targeting developers interested in AI chatbot integration.

2. **Method of Attack**: The packages were published on PyPI by a user named “Xeroline” and claimed to provide API access to OpenAI’s GPT-4 Turbo and Anthropic’s Claude. However, they only offered a limited interaction with a free demo version, while delivering malware covertly.

3. **Infection Strategy**: Publishing malicious open-source packages is a method employed by cybercriminals to compromise application developers and their organizations, exploiting their eagerness for free access to popular services like AI tools.

4. **Infostealer Capabilities**: JarkaStealer can steal sensitive data, take screenshots, and capture session tokens from applications such as Telegram, Discord, and Steam. It is marketed on the Russian Dark Web for $20, with various modifications available.

5. **Duration and Impact**: The malicious packages were available on PyPI for approximately one year and were downloaded over 1,700 times across multiple countries, predominantly in the United States, before being taken down by Kaspersky’s researchers.

6. **Download Statistics**: Initial download numbers may have been artificially inflated by the attacker to create a façade of legitimacy. A significant drop in downloads after the first day indicates manipulation of download statistics.

7. **Security Recommendations**: Security professionals advise users to check the popularity and legitimacy of Python packages before downloading. However, the average user may not take these precautions, making them vulnerable to such attacks.

8. **Caution Advised**: These incidents highlight the importance of exercising caution when dealing with open-source packages and the potential for cyber threats in pursuing access to high-demand tools like AI platforms.

Full Article