Privacy threat? New Amazon patent means Alexa could always listen and record every word spoken to it
Amazon is at the centre of a privacy scare after the e-commerce giant filed a patent that would allow Alexa to record everything a person says, before a command is actually issued.
This patent is sure to trigger privacy concerns, as it could mean that Alexa equipped devices would constantly record audio around the smart speaker.
Amazon is already facing privacy questions after it was reported last month that a global team of people at Amazon reviewed audio clips of people speaking to their Alexa-powered smart speakers.
The teams apparently review the recordings in order to improve Alexa’s understanding of human speech and make it better at responding to commands.
These Amazon teams are said to span many sites around the world, such as in Boston (US), through to Costa Rica, India and Romania. The team is said to number in the thousands, and is made up of both contractors and full-time Amazon employees.
But now Amazon is seeking to move one stage further after the discovery of a recent patent application, entitled ‘pre-wakeword speech processing.’
Amazon describes it as “a system for capturing and processing portions of a spoken utterance command that may occur before a wakeword.”
“The system buffers incoming audio and indicates locations in the audio where the utterance changes, for example when a long pause is detected,” it said. “When the system detects a wakeword within a particular utterance, the system determines the most recent utterance change location prior to the wakeword and sends the audio from that location to the end of the command utterance to a server for further speech processing.”
Essentially it means that if the new patent is implemented, Alexa-devices would constantly record audio around them, allowing them to respond to a command even if the wake word is at the very end.
It should be remembered that at the moment smart speakers powered by Alexa artificial intelligence, continuously listens to ambient sounds, but only sends recordings to be stored on Amazon’s servers following a “wake word” such as “Echo”, “Amazon” or “Alexa”.
But the device can sometimes mishear words and sounds, and mistakenly send recordings to the cloud.
The new Amazon patent would effectively allow users to add the Alexa wake word to the end of their conversations, and not at the start of them, and it would still respond.
So for example, “Alexa, will I need coat tomorrow?” could be replaced with “Will I need a coat tomorrow, Alexa?”
This means Alexa will always be listening and possibly recording – an issue that has not gone unnoticed by security experts.
“Many Amazon Alexa users will likely be alarmed by today’s news that the company’s latest patent would allow the devices – commonplace in homes across the UK – to record everything a person says before even being given a command,” said David Emm, principal security researcher at Kaspersky Lab.
“Whilst the patent doesn’t suggest it will be installed in future Alexa-enabled devices, this still signals an alarming development in the further surrender of our personal privacy,” said Emm. “Given the amount of sensitive information exchanged in the comfort of people’s homes, Amazon would be able to access a huge volume of personal information – information that would be of great value to cybercriminals and threat actors. If the data isn’t secured effectively, a successful breach of Amazon’s systems could have a severe knock-on effect on the data security and privacy of huge numbers of people.”
Emm went onto to warn that consumers need to be informed of the change and its implications.
“If this patent comes into effect, consumers need to be made very aware of the ramifications of this – and to be fully briefed on what data is being collected, how it is being used, and how they can opt out of this collection,” said Emm. “Amazon may argue that analysing stored data will make their devices smarter for Alexa owners – but in today’s digital era, such information could be used nefariously, even by trusted parties. For instance, as we saw with Cambridge Analytica, public sector bodies could target election campaigns at those discussing politics.”
“There’s a world of difference between temporary local storage of sentences, to determine if the command word has been used, and bulk retention of data for long periods, or permanently – even if the listening process is legitimate and consumers have opted in,” concluded Emm. “There has already been criticisms of Amazon for not making it clear what is being recorded and stored – and we are concerned that this latest development shows the company moving in the wrong direction – away from data visibility, privacy, and consent.”
There have been a number of issues with Echo speakers since their arrival in June 2015.
Last Christmas for example, one case reportedly had Alexa telling a user to “kill your foster parents”.
Amazon eventually found that the line had been taken out of context from a Reddit discussion.
But Amazon has also been dragged into legal cases, including murder trials.
Last year judge in the US asked the firm to hand over audio recordings from an Amazon Echo.
The Echo speaker in question was located in a house in New Hampshire, where the bodies of two women with multiple stab wounds were found under the porch of the house.
Prior to that in 2016, US police wanted access to some audio data that may have been recorded on an Amazon Echo in a murder investigation of Victor Collins, who was found on 22 November 2015, in a hot tub at the home of Andrew Bates.
Police found signs of a struggle, including spots of blood and broken bottles, as well as an Amazon Echo on the home’s kitchen counter.
In March 2017 the impasse was resolved when the defendant, Andrew Bates, consented to allow the data to be handed over, which Amazon did.
Think you know all about Amazon? Try our quiz!