Last week, Ava Kofman wrote an interesting yet terrifying piece in The Intercept about Voice RT. You've probably never heard of Voice RT before because it's been one of those things the U.S. Government does in secret; in this case, it was developing technology that can positively identify someone by the sound of their voice. Be sure to read it. It's important information everyone needs to know.
Our voice is the perfect biometric identifier; it's stable and unique.
Now the idea of recognizing someone by their voice is easy to grasp. We do it every day when we talk to the people close to us. A person's voice is pretty unique and it doesn't take a lot of processing power — either the organic kind in our heads or the silicon kind in our gadgets — to know who you're talking to just from hearing them speak. But the NSA was able to take things to the extreme. They have the ability to listen to everything, everywhere. If you're using a pay phone in the middle of nowhere they can listen. They might even have the authority to do it, and that means we might have a big problem sitting on our nightstands or coffee tables called Google Home and Amazon Echo.
Speaking with The Verge, Albert Gidari, director of privacy at the Stanford Center for Internet and Society, said these products "are vulnerable to government demands for access and disclosure; I think the government could obtain a technical assistance order to facilitate the scan, and under FISA, perhaps to build the tool, too." I'm sure Gidari is right because the government can already subpoena our phones, our computers, and even our televisions to facilitate an investigation. Senators Wyden (D., Oregon) and Paul (R., Kentucky) think that FISA might be abused in this way, too.
And it's not an issue of a government agency spying on what we are saying or doing when we talk to Alexa or Google Home. They don't need that information; all they want is a recording of our voices.
Both the Echo and Google Home store recordings of the things we say when talking to them. Both also transmit voice data to a cloud server for processing, too. Thankfully, each only records after the hot word is detected and any data that leaves your device is encrypted and anonymized. Whether they could intercept and decrypt the data, demand it through the courts, or pull it right from the device all they'll hear is people like you and me telling Assistant or Alexa to change the channel or give a weather forecast — mundane information that you aren't trying to keep secret. Police in Bentonville Arkansas found just that when an Amazon Echo was subpoenaed in a 2016 murder case. But again, that's not what the NSA was collecting with Voice RT; they just wanted some sample data of a voice so they could match it while doing the spying thing in real time.
It's not what you say. It's the fact that it is you saying it that makes it interesting to the NSA.
Digging through the documents leaked by Edward Snowden, Kofman found that the NSA has been collecting voice recognition data for years. This technology was used to identify Saddam Hussein and match some past recordings that were captured. Voiceprints for Osama Bin Laden and other high-ranking Al Qaeda members were created, and one memo tells how Abu Musab al-Zarqawi was found to be the speaker in online audio files that the CIA was very interested in. These classified documents show that between 2004 and 2012 the NSA refined and used their speaker recognition technology in counterterrorism ops and international drug arrests.
If you think the country's top spy agency collecting random recordings of what people say to assist when fighting terrorism or drug trafficking is OK, not many people would argue with you. Google and Facebook (and probably every other internet company) scan collected data for things like child pornography or abuse, copyright infringement, and terroristic activity because someone feels it's for the greater good. I won't argue. But Snowden also revealed that the NSA plans to deploy the same tech to prevent whistleblowers like him from exposing their misdeeds years before Executive Order 13587 was signed.
The NSA has plans to use the tactics that help catch Bin Laden to surveil Americans.
That's going too far, and it has people like former White House adviser to the Director of National Intelligence Timothy Edgar worried about our privacy and the repercussions from an overbearing government agency if they think we have said or done something they don't like. We've all heard cases where a very thin line was walked and crossed and everything wasn't quite legal according to the spirit of the Constitution. When it comes to our voices, civil liberty experts think Google and Amazon need to change how Home and the Echo work so that no voice data is retained.
Our voices are stable; something I said 10 years ago can be positively identified through a voiceprint as coming from me today. With voice recordings at hand, even though they are files saying something like "OK Google, dim the living room lights to 50%" it's easy to use them against us for any investigation or operation. I imagine the NSA has plenty of voice data that can identify a person who served drinks to a wanted terrorist at a hotel, or the agent at the ticket counter in an airport. I also imagine these innocents were questioned and then served with a gag order, even though they were innocent of any crimes and the NSA and CIA and FBI knew they were innocent.
Don't believe the talk that programs like FISA are designed and used solely to protect us. We have plenty of evidence to the contrary.
This needs to be unmasked and debated in public. Technology that can be used to track journalists or expose their sources is dangerous when not used correctly and shouldn't be hidden behind a wall of government classification. While groups like the EFF and Freedom of the Press Foundation try to make that happen, there is pressure on Google and Amazon to stop saving and analyzing voice data the way they do now. That puts quite the damper on a system designed to analyze everything and get smarter. Machine learning needs copious amounts of data to analyze over and over to "learn." Google and Amazon need to worry about collecting and storing it all in a way that keeps it anonymous instead of worrying about how their learning machines can learn without the data they need.
A machine needs data to learn. A lot of data repeated over and over and over again.
Pointing the finger at Alexa and Assistant as the root of the problem isn't good for anyone who finds them useful. We need to focus on roping in our government and putting a stop to unwarranted surveillance of Americans. And this isn't just a U.S.A. thing; Interpol and Britain's GCHQ have "worked closely" with the NSA and credit programs like Voice RT as "playing an important part in our relationship with NSA." China has been said to have the same type of program and is now able to positively identify tens or perhaps hundreds of thousands of Chinese citizens by the sound of their voice automatically.
I don't trust the NSA to have secret technology that can identify any of us at any time and not use it inappropriately. I agree with civil liberty experts that this is a dangerous path and should be made public. I don't agree with pressuring Google and Amazon to stop innovating while we wait. We're seeing the future unfold before our eyes and until the "smart" machines Google and Amazon are using to bring it to us show that they need reigning in, let's not hold the future back.
