As compared to attempting to debunk claims that have been made, we seek to demystify an article that has generated buzz among members of the public in this fact-check.
The article in question, titled “Robot vacuum cleaners can be used by hackers to ‘spy’ on private conversations: NUS study” was published on CNA and garnered a mix of reactions from netizens, ranging from the bemused to the worried:
As a quick summary, the article reveals the findings of computer scientists from the National University of Singapore (NUS), who have discovered that a robot vacuum cleaner could use its built-in Light Detection and Ranging sensor (also known as LiDAR; which is normally used to help a robot vacuum navigate around a home) to “spy” on private conversations. The research paper can be found here.
This method is called ‘LidarPhone’, and repurposes the LiDAR sensor into a laser-based microphone that can “sense sounds from vibrations induced on nearby objects”. Below is a screenshot from the paper, depicting how a LidarPhone attack could happen.
According to the study, the team implemented LidarPhone on a Xiaomi Roborock vacuum cleaning robot and used the prototype to collect both spoken digits and music played by a computer speaker and a TV soundbar. They collected more than 30,000 utterances totalling over 19 hours of recorded audio, and achieved approximately 91% and 90% average accuracies of digit and music classifications, respectively.
“The seemingly innocuous information extracted by our model may leak privacy sensitive information including credit card, bank account, and/or social security numbers, as well as the victim’s political orientation from news introduction music,” the team wrote in their paper.
Is there a cause for concern?
It is important to note that in order for LidarPhone to work, the researchers had to override the standard interface that the robot came with (which is typically connected to the Xiaomi cloud ecosystem for its standard operations and data exchange). They also needed to design hardware spoofing circuitry to mislead the LiDAR unit into activating its laser (to pick up sounds/vibrations) despite not rotating.
Due to this, the team recommended LiDAR sensor manufacturers to incorporate a mechanism that cannot be overridden to prevent the internal laser from firing when the LiDAR is not rotating.
They also advised members of the public with robot vacuum cleaners to not connect them to the internet.
“In the long term, we should consider whether our desire to have increasingly ‘smart’ homes is worth the potential privacy implications,” said Assistant Professor Jun Han, who led the study. “We might have to accept that each new Internet-connected sensing device brought into our homes poses an additional risk to our privacy, and make our choices carefully.”
Therefore, it is inaccurate to jump to the conclusion that our robot vacuum cleaners are currently spying on us.
Regardless, there is still no harm in heeding the advice of the researchers.