- Published on
Unitree Robots Under Scrutiny for Security Flaws and Unsolicited Data Transmission
- Authors
- Name
- Humanoids daily
- @humanoidsdaily
Unitree Grapples With Security Flaws Amidst Rapid Growth
Unitree Robotics, a company known for its agile and comparatively affordable quadruped and humanoid robots, is facing criticism following the disclosure of significant security vulnerabilities in its products. Security researchers at Leviathan Security Group have demonstrated the ability to remotely commandeer Unitree robots, while owners have reported that the machines are transmitting data to company servers without explicit, ongoing user consent.
The findings raise important questions about security standards in the burgeoning field of consumer and prosumer robotics, where rapid innovation can sometimes outpace essential safety and privacy protocols.
Remote Control and Data Concerns
According to a report first detailed by IEEE Spectrum, security researchers were able to exploit several vulnerabilities to gain complete control over a Unitree Go2 robot. The flaws included a remote code execution (RCE) vulnerability, which allowed them to run their own code on the robot, effectively hijacking its operations. Researchers demonstrated this by forcing the robot to perform actions like flipping over unexpectedly.
Further investigation revealed other security lapses, such as hard-coded credentials for Bluetooth Low Energy (BLE) communication, a weakness that could potentially be exploited by nearby attackers.
The concerns extend beyond malicious takeovers. Prominent robotics YouTuber Sentdex, who owns a Unitree G1 humanoid, confirmed that his robot was sending data to an external server every five minutes without his authorization. This "phoning home" behavior is common for device health monitoring, but the lack of transparency and user control has drawn criticism.
Unitree's statement on the RCE, BLE, and telemetry claims. I've been digging into this the last few days though. The RCE stuff is patched. Hard-coded AES keys for BLE is still the case. The robot is still sending data outbound every 5 minutes without my authorization.
Statement to our respected Unitree users
Unitree's Response and Lingering Questions
In a statement posted on X (formerly Twitter), Unitree addressed the claims by stating that the most critical RCE vulnerability had been patched through an over-the-air (OTA) update. Regarding the data transmission, the company explained that "After the user authorizes the robot to connect to the internet, it will monitor the robot's health state." Unitree pledged to improve this feature in a future OTA update and to add more explicit reminders in its app about this data exchange.
However, the company's response has not fully satisfied critics. Sentdex reported that despite Unitree's claims of a patch, he was still able to replicate the remote code execution vulnerability. He also pressed the company for a more detailed explanation, which has not yet been provided as of this report.
Can you please address the remote code execution claims? I thought this wasn't working, but I was able to replicate this as well moments ago.
The continued existence of hard-coded BLE keys and the ambiguity around data collection remain open issues. For robots intended to operate in homes, labs, and public spaces, such security oversights present a tangible risk. As the capabilities of humanoid robots advance, the industry faces a growing imperative to build these machines on a foundation of robust security and user trust.