Snooping microphones in your home?

[This is an ABC podcast]

Dr Karl: George Orwell’s book, 1984, describes a world which is in a state of continual war and which has an oppressed population kept in line by government surveillance wherever they go. The omnipresent “telescreens” not only broadcast propaganda from Big Brother and fake news 24 hours a day – but they also use their video cameras and microphones to snoop on citizens.

Today, just like in 1984, many of us have microphones in our homes. But the difference is that we voluntarily purchase them. These microphones are embedded in Smart Devices and are activated by software or Digital Assistants such as Amazon’s Alexa, Apple’s Siri and Google’s Assistant.

Today, around 22 per cent of Americans use voice recognition devices such as Amazon Echo and Google Home. These devices form part of the Internet of Things that many of us have entered into, where our toasters, fridges, sound systems, house lights, and even our front door locks are all linked into the Interwebs.

Now, sure it is convenient to be able to say, “Hey Siri, three-minute countdown” when you want to time the boiling of an egg and you don’t want to wipe your food-covered hands all over your nice clean Smart Phone. But some people are worried that these Digital Assistants are snooping on us when we don’t expect them to. The companies selling them do insist that their devices record what they hear only after they have been deliberately triggered by a specific voice command.

But that hasn't always been how it's worked in practice. In 2017, a tech journalist found that his Google Home Mini was recording him almost 24/7, thanks to a glitch in the device. That problem was fixed quick-smart, but it's not the only example. In another case, Arkansas police investigating a possible murder got access to voice recordings captured by an Amazon Echo in 2016. They thought the recordings might help them work out how the person had died. The police got the recordings with the permission of the defendant.

Delving into the patents of companies like Amazon and Google is fascinating too. In one of its patents, Amazon describes what it calls a “Voice Sniffer Algorithm.” This would let the Amazon digital assistant listen into a phone call, monitor the conversation for words such as “like,” “bought” or “dislike” – and a little while later, you could get a personalised advertisement for the San Diego Zoo if you had talked about a fluffy puppy or a Wine of the Month Club pop-up ad on your phone or computer the next time you cruised the web if, say, you had mentioned that you liked wine.

On the one hand, Amazon says it will not use this technology for personal advertising … but on the other hand the patent itself explicitly mentions that it could be used for "targeted advertising and product recommendations." Hmmm.

Meanwhile, Google has filed a Patent Application to sample a child’s speech patterns, whispers or silence to see if they were engaging in "mischief." Sounds a bit like a baby monitor on steroids.

The non-profit advocacy group, Consumer Watchdog, studied many of these patent applications and wrote, “ … it’s really clear that this is spyware and a surveillance system meant to serve you up to advertisers.”

So our devices are listening to us, and sometimes it's when we don't want them to. But what about when we talk to them? It's meant to be a special relationship … that your intelligent assistant will obey your command alone. But there’s a catch. It seems that other unauthorized people can secretly talk to our various digital devices and get them to do stuff that we don’t want – such as add items to a shopping list.

Back in 2016, students from Georgetown University and the University of California were able to hide surreptitious commands in white noise - that came either through loudspeakers in your house, or YouTube videos. These commands could make your phone go into aeroplane mode or even open a website.

By 2018, the students from the University of California successfully embedded commands into spoken voice (such as a podcast) and recordings of music – and these commands were completely unnoticed by the owner of the phone or the device.

Another group of researchers from China’s Zhejiang University and Princeton University were able to activate people’s voice recognition systems via sounds that the human ear could not hear. In response, the Smart Devices made phone calls, took pictures, sent texts, visited dangerous websites and even got into Smart Home Accessories such as your front door lock. Now you can imagine the possibilities for cybercrime.

By the year 2021 in the USA there will be more phones and loudspeakers with digital assistants than humans. At the moment, Apple’s Siri, Google’s assistant and Amazon’s Alexa can all be tricked into obeying commands that we humans cannot detect, and which are embedded into voice or music. Mind you, the companies will fix that, and the baddies will invent something new.

The technology is moving so quickly that it’s difficult for the average non-techo consumer to protect themselves. And the tech companies have to balance security against ease-of-use by the consumer. Maybe it’s time for a return to sign language and invisible ink, or maybe telepathy will be our saviour. Most of all, I really hope in the future I won't be saying – 'Big Brother, set an egg timer for three minutes.'

Computerised voice: Dr Karl.

Dr Karl: Who's that?!

Computerised voice: It's Big Brother here. As you requested, I'm setting a timer for three minutes. Don't forget … I'm always watching you. And listening! I just can't stop listening to your podcast, Great Moments in Science.

Dr Karl: Aww, thanks Big Brother. I think.


© Australian Broadcasting Corporation