<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div class="container font-size5 content-width3">
      <div class="header reader-header" style="display: block;"> <font
          size="-2"><a class="domain reader-domain"
href="https://theintercept.com/2018/06/25/interpol-voice-identification-database/">https://theintercept.com/2018/06/25/interpol-voice-identification-database/</a></font>
        <h1 class="reader-title">Interpol Rolls Out International Voice
          Identification Database Using Samples From 192 Law Enforcement
          Agencies</h1>
        <div class="credits reader-credits">Ava Kofman - June 25, 2018<br>
        </div>
      </div>
      <hr>
      <div class="content">
        <div class="moz-reader-content line-height4" style="display:
          block;">
          <div id="readability-page-1" class="page">
            <div>
              <p><u>Last week, Interpol</u> held a final project review
                of its speaker identification system, a four-year, 10
                million euro project that has recently come to
                completion. The Speaker Identification Integrated
                Project, what they call SiiP, marks a major development
                in the international expansion of voice biometrics for
                law enforcement uses — and raises red flags when it
                comes to privacy.</p>
              <p>Speaker identification works by taking samples of a
                known voice, capturing its unique and behavioral
                features, and then turning these features into an
                algorithmic template that’s known as a voice print or
                voice model. With enough voice prints and samples
                collected in its global audio database, Interpol’s
                speaker identification system will be able to upload an
                unknown voice and, regardless of the language it is
                speaking, match it to a list of likely candidates.
                SiiP’s database allow uploads and downloads of samples
                from 192 law enforcement agencies across the world.</p>
              <p>SiiP will join Interpol’s existing fingerprint and face
                databases, and its key advantage will be to facilitate a
                quick identification process — say, of a kidnapper
                making a phone call — even in the absence of other
                identifiers. The platform also boasts the ability to
                filter voice samples by gender, age, language, and
                accent. When the audio recordings are taken from similar
                acoustical environments, accuracy rates can be extremely
                high.</p>
            </div>
            <div>
              <p>Speech recognition technologies can identify and tag
                individuals every time they open their mouths,
                effectively ending <a
href="https://www.forbes.com/sites/realspin/2016/10/06/voice-recognition-every-single-day-every-word-you-say/#30bc3f48786d">anonymity</a>.
                As Interpol’s <a href="https://vimeo.com/219362794">promotional
                  video explains</a>, departments using SiiP can upload
                intercepted phone calls and also search against voices
                on social media. SiiP’s database will include samples
                from YouTube, Facebook, publicly recorded conversations,
                voice-over-internet-protocol recordings, and other
                sources where individuals might not realize that their
                voices are being turned into biometric voice prints.
                “People choose to upload material online for various
                reasons, but I doubt it’s to let police and arms
                companies then enroll them into secret databases made
                available to police around the world,” explained Edin
                Omanovic, a surveillance expert at Privacy
                International.</p>
              <p>Cynthia Wong, a researcher at Human Rights Watch,
                warned that a broad mandate could lead to an
                ever-expanding collection. “There are many instances
                where we might consent to our voice being recorded for
                one purpose, but would object to using it for others,
                including using our voice to build and train a massive
                voice biometric database and recognition system,” she
                said. “Or perhaps we didn’t consent to our voice being
                recorded at all — perhaps our voice was secretly
                recorded or inadvertently caught in the background of a
                recording, but has now been placed in Interpol’s
                database.</p>
              <blockquote data-shortcode-type="pullquote"
                data-pull="none">SiiP’s database will include samples
                from YouTube, Facebook, publicly recorded conversations,
                and other sources where individuals might not realize
                that their voices are being turned into biometric voice
                prints.</blockquote>
              <p>Interpol’s system appears to represent the largest
                international voice-harvesting effort to date. But as a
                recent Interpol survey reports, many law enforcement
                agencies have already been quietly using voice
                recognition systems. More than half of the 91
                departments in 69 countries surveyed already run
                automated speaker identification programs. In 2010,
                Mexico <a
                  href="https://speechpro-usa.com/media/news/2010-06-03">announced</a>
                that it had created the world’s first nationwide
                automated speech recognition system, using Russia’s
                Speech Technology Center. As The Intercept <a
href="https://theintercept.com/2018/01/19/voice-recognition-technology-nsa/">reported</a>
                in January, the National Security Agency has been using
                speaker recognition to monitor terrorists, politicians,
                drug lords, spies, and agency employees since at least
                2004. The Department of Homeland Security’s new HART
                database — which already stores a fraction of the world
                population’s biometrics — will support voice data, in
                addition to DNA, scars, and tattoos. Human Rights Watch
                released a bombshell <a
href="https://www.hrw.org/news/2017/10/22/china-voice-biometric-collection-threatens-privacy">report</a>
                last year exposing China’s state-of-the-art system,
                which appears to be integrated with mobile phone
                networks and recognizes the voices of known suspects in
                real time. Wong says that China’s integrated
                surveillance system is likely to set a precedent for
                other countries to follow suit.</p>
              <p>It is unclear what kinds of legal safeguards exist to
                protect against the creation of international biometric
                collections, which may retain data indefinitely.
                Oversight bodies for intelligence sharing are <a
href="https://privacyinternational.org/feature/1742/new-privacy-international-report-reveals-dangerous-lack-oversight-secret-global">scarce</a>.
                “How can Interpol ensure that the voice samples other
                agencies submit are lawfully intercepted?” Wong asked.
                “Based on our work, many countries around the world fall
                woefully short in regulating surveillance to begin
                with.”</p>
              <p>Wong, Omanovic, and others are also concerned about the
                plans to create a “blacklist” of criminal and suspect
                voices. It is not known who will audit the list, what
                rules will govern when and how countries add to it, and
                what happens if someone is placed on the blacklist in <a
href="https://www.scientificamerican.com/article/voice-analysis-should-be-used-with-caution-in-court/">error</a>,
                Omanovic said. Wong wonders what kinds of safeguards
                will be used in countries where the government regularly
                criminalizes dissent or jails journalists.</p>
              <p>An Interpol report researching the ethics of biometrics
                acknowledges the great chilling effect that such
                collection might have. “Processing of audio data that is
                never actually viewed does not count as intrusive in
                itself, though it can lead to future intrusion,” it
                states. “Particularly in an investigatory context where
                the aim is identifying suspects for future scrutiny.
                Even when no human viewing does take place, the
                application of these techniques has consequences
                resembling genuine intrusion: They can easily cause
                people to fear that their information will be exposed,
                creating a harmful incentive to avoid associational life
                or unconventional behavior.”</p>
              <p>Verint, a multinational biometrics company, has been
                leading project development, and did not respond to a
                request for comment. As Privacy International has
                documented, the company has been known to <a
href="https://privacyinternational.org/press-release/1186/privacy-international-uncovers-widespread-surveillance-throughout-central-asia">install</a>
                and “<a
href="https://privacyinternational.org/blog/1362/how-colombia-built-shadow-state-new-privacy-international-investigation-reveals">sell</a>
                intrusive and mass surveillance systems worldwide
                including to authoritarian governments.” This, according
                to Omanovic, “tells you everything you need to know
                about the level of ethical scrutiny being applied. The
                EU is being securitized without any public debate, and
                the only people who will ultimately benefit are those
                working in the arms industry.”</p>
            </div>
          </div>
        </div>
      </div>
      <div> </div>
    </div>
    <div class="moz-signature">-- <br>
      Freedom Archives
      522 Valencia Street
      San Francisco, CA 94110
      415 863.9977
      <a class="moz-txt-link-freetext" href="https://freedomarchives.org/">https://freedomarchives.org/</a>
    </div>
  </body>
</html>