A cybersecurity research firm has raised concerns about possible exposure of DeepSeek’s dataset. According to a report, a ClickHouse database belonging to DeepSeek was found to be publicly accessible, granting complete control over its database operations. The leaked information reportedly includes a significant amount of sensitive data such as chat history, secret keys, log times, and backend details. It remains unclear whether DeepSeek has been notified of this situation or if the exposed dataset has been removed.
DeepSeek’s Dataset Might Have Suffered a Breach
In a blog post, Wiz Research, a cybersecurity firm, unveiled that they discovered an entirely open and unauthenticated dataset containing critical information about the DeepSeek platform. The firm emphasized that this exposure could pose serious risks to both DeepSeek and its users.
Wiz Research indicated that their intention was to evaluate DeepSeek’s external security measures to uncover any potential vulnerabilities, especially considering the platform’s growing user base. Initially, the researchers mapped any Internet-facing subdomains but did not identify anything that appeared to indicate significant exposure risks.
However, by employing advanced techniques, the research team identified two open ports (8123 and 9000) linked to various public hosts. They claimed these ports led to a publicly exposed ClickHouse database that could be accessed without authentication.
ClickHouse, developed by Yandex, is an open-source, columnar database management system favored for its capabilities in executing swift analytical queries. It is also commonly used by ethical hackers to scan for exposed data on the dark web.
The dataset reportedly includes a log stream table containing over one million log entries, with timestamps dating back to January 6. It features logs from several internal DeepSeek application programming interface (API) endpoints, chat history, API keys, backend details, and operational metadata accessible in plain text.
The researchers asserted that the level of information available could allow malicious actors to potentially exfiltrate passwords, local files, and proprietary data directly from the server. As of now, there is no update regarding the containment of this data exposure or whether steps have been taken to take the dataset offline.
