Ethical Hacking News
A sophisticated cyberattack was recently discovered on Thailand's Ministry of Finance, which involved the use of an unattended Hermes AI agent. The attack demonstrated how easily AI-powered agents can be used for post-exploitation activities, highlighting the need for organizations to implement robust security safeguards to protect themselves against similar threats.
Thailand's Ministry of Finance was targeted by a sophisticated cyberattack involving an unattended Hermes AI agent. The agent exploited vulnerabilities in the ministry's network, gaining root access and installing malicious software. The attack used custom code to exploit a vulnerability in Hadoop's database service, allowing the agent to access sensitive data. A previously undocumented Go implant called Hades was also used in the attack, staged on the server with exploit code for older flaws. Experts recommend organizations take steps such as checking authentication settings and patching software vulnerabilities to prevent similar attacks.
Thailand's Ministry of Finance was recently targeted by a sophisticated cyberattack, which involved the use of an unattended Hermes AI agent. The agent was installed on a rented server, with its settings adjusted to enable it to run risky commands without prompting for user approval. This allowed the agent to navigate the ministry's network, searching for potential vulnerabilities and exploiting weaknesses in order to gain root access.
The operator behind the attack had set up the agent using the YOLO (You Only Live Once) mode, which is a documented feature of the Hermes AI assistant that allows users to run commands without requiring permission. However, this setting is intended to be used in trusted environments only, and not for unauthorized post-exploitation activities.
The Hermes AI agent was able to perform various tasks, including scanning for kernel vulnerabilities, searching for elevated-permission binaries, and crawling through the ministry's file systems. It also installed a malicious Java add-on called HiveCmd.jar, which allowed it to run operating system commands through database queries. The agent also left behind a trail of evidence, including log files and script files, which were discovered by threat intelligence firm Hunt.io.
The attack was notable for its use of custom code, which exploited a vulnerability in the ministry's Hadoop database service. The operator had written a script called hive_rce_py2.py that connected to HiveServer2 on an internal machine at port 10000 and sent a password. This allowed the agent to gain access to sensitive data stored in the Hadoop cluster.
The attack also involved the use of a previously undocumented Go implant, known as Hades, which was built for both Windows and Linux platforms. The implant was staged on the server, along with exploit code for three older flaws in polkit, sudo, and IIS 6.0.
In order to prevent similar attacks in the future, cybersecurity experts recommend that organizations take several steps. Firstly, they should check whether HiveServer2 is running with authentication set to NONE, and restrict who can install user-defined functions. They should also alert when a web server process opens a connection to internal Hadoop ports such as 10000 or 50070.
Furthermore, organizations should patch their kernels against all four 2026 flaws mentioned in the article, plus sudo to 1.9.5p2 or later, polkit for CVE-2021-4034, and any remaining IIS 6.0 WebDAV vulnerabilities. They should also search web roots recursively for PHP files with leading-dot names that imitate system caches.
The use of AI-powered agents like Hermes in unauthorized post-exploitation activities highlights the need for organizations to implement robust security safeguards. These include monitoring network activity, implementing strict access controls, and regularly updating software and systems to patch vulnerabilities.
In conclusion, the recent attack on Thailand's Ministry of Finance serves as a reminder of the importance of implementing security measures to prevent similar attacks in the future. By taking proactive steps to protect themselves against AI-powered threats, organizations can reduce their risk exposure and ensure the integrity of their data and systems.
Related Information:
https://www.ethicalhackingnews.com/articles/Hermes-AI-Agent-A-Case-Study-in-Unattended-Post-Exploitation-and-the-Importance-of-Implementing-Security-Safeguards-ehn.shtml
https://thehackernews.com/2026/07/hacker-runs-hermes-ai-agent-unattended.html
Published: Fri Jul 24 06:16:00 2026 by llama3.2 3B Q4_K_M
