Category
Robot security research
Website
darknavy.org
Last verified
November 1, 2026

What is DARKNAVY?

DARKNAVY describes itself as an independent cybersecurity research and services institute with locations in Singapore and Shanghai. It traces its lineage to KeenTeam, founded in 2011, and officially launched in 2023. Its founding team runs the GEEKCON hacking competition and created the AI security product deepsec and the Adversarial Vulnerability Scoring System (AVSS). The institute offers in-depth security assessments of products and systems for enterprises, and publishes research on operating systems, chipsets, AI, mobile, IoT and connected hardware.

Key capabilities

  • Product and system security assessments for enterprises, based on original vulnerability research.
  • Research across many layers: operating systems, chipsets, AI, mobile, IoT and Web3.
  • AI-driven vulnerability discovery, including its deepsec product and a multi-agent research architecture it calls Argusee.
  • Hardware teardown research, with published work on devices such as drone controllers and satellite user terminals.
  • GEEKCON, a hacking competition run by its founding team where new attacks are demonstrated live.

Why it matters for humanoid robot security

Editorial analysis: DARKNAVY is one of the few research teams with published, hands-on work on commercial legged robots. Its March 2025 write-up "The Jailbroken Unitree Robot Dog" analyzed root access on the Unitree GO2, including unauthenticated local WebRTC access and a file-upload path that could overwrite a built-in script (removed in later firmware). The team also said an earlier GO1 issue it reported to Unitree went unanswered. Press coverage of GEEKCON in Shanghai described DARKNAVY researchers hijacking a Unitree robot through voice commands aimed at its AI agent, with the compromise then spreading wirelessly to a nearby robot. Humanoids from the same vendors share much of this software stack, so this kind of research is an early warning for humanoid buyers.

Where it fits in a humanoid robot security program

Analysis: DARKNAVY's published work sits on the robot itself: local network services, firmware and the onboard AI agent that turns speech into actions. In a humanoid program that makes it most relevant at the testing stage, before a fleet is bought or shipped, when an independent team can probe wireless setup, remote-control channels and voice or LLM interfaces. Its disclosure history is also a useful input at procurement, when buyers check how a vendor handled past reports.

Who it is for

  • Robot makers that want independent offensive research on their legged or humanoid platforms
  • Enterprises in Asia-Pacific evaluating Chinese-made robots before deployment
  • Security teams tracking real-world robot exploitation techniques

Questions to ask DARKNAVY

  • Which robot platforms have you assessed, and can you share redacted findings?
  • Do your assessments cover the robot's AI agent and voice interface, not just network services?
  • How do you coordinate disclosure with robot vendors that do not respond?
  • Can you retest after the vendor ships a firmware fix?

Similar listings

Sources

This listing is editorial and free; inclusion is not an endorsement. Work at DARKNAVY? Claim or update this listing.

Get the weekly Humanoid Threats Brief → Subscribe