A.I. See the Future of Fraud

  • +5
    Resident47
    | 1 reply
    Probably anyone reading here caught the impostor call story from New Hampshire a couple weeks ago, when some primary election disruptor thought to blast the state with a simulated President Biden and its attempt at social sabotage. NBC News is credited with the story break, and aired the longest audio clips. PBS News Hour I think did much better than most in describing the event and its background and why the call was so misleading:

    AI robocalls impersonate President Biden in an apparent attempt to suppress votes in New Hampshire - PBS, 22 Jan 2024

    A notable side mission of the call sweep was to paint a target on a high profile Biden supporter:
        "... they falsely showed up to recipients as coming from the personal cellphone number of Kathy Sullivan, a former state Democratic Party chair who helps run Granite for America, a super-PAC supporting the Biden write-in campaign.

    Many reports were frustratingly inarticulate on the call delivery, and those which ran captured audio offered only excerpts of varying length. Some said there was a keypress option "to be removed", others said that Ms. Sullivan's phone number was given, and it was not clear if that meant within the audio delivery or the Caller ID or both.

    After a good deal of search spelunking with different transcript quotes, I finally isolated the PAC's number and a direct call capture on one of the call blocker services. What follows is the most complete transcript thus far, from a bot call in progress as the capture began:

        ... the New Hampshire presidential preference primary. Republicans have been trying to push nonpartisan and Democratic voters to participate in their primary. What a bunch of malarkey. We know the value of voting Democratic when our votes count. It's important that you save your vote for the November election. We'll need your help in electing Democrats up and down the ticket. Voting this Tuesday only enables the Republicans in their quest to elect Donald Trump again. Your vote makes a difference in November, not this Tuesday. If you would like to be removed from future calls, please press two now. [Biden voice ends] Call 603-xxx-nnnn to be removed from future calls.

    The final line citing the phone number abruptly switches to an appended track, performed I believe by a human. It's in some other man's voice in a southern twang, Texas flavored at a guess. Maybe this alternate removal method was pasted on as an afterthought, obviously not for the stated purpose of hit list deletion. Granite for America has been apparently publicizing that phone number since 2017 at least. Still I've opted to 'sanitize' it here to avoid inspiring any more crank 'n' prank callers. I'm going to assume that Caller ID also spoofed the same number to ramp up the damage.

    News reports also have varied in describing how convincing the Biden-bot track sounded, often noting the observant "malarkey" touch. Not everyone has the same ear for mimicry. My early impression was a cut-paste job of acquired clips, but the crashed edits and clunky pausing and overall vacuous delivery all lean toward an algorithmic actor. Bogus Biden sounds like a hostage reading his captor's prepared speech, very much out of character. I came to realize that the garden variety fraud would be too lazy to assemble an audio montage, preferring to plug his script into some AI bucket and release the raw output with little to no cleanup. An analyst contacted by PBS seemed to agree:
    Quote:
        “We have been concerned that generative AI would be weaponized in the upcoming election and we are seeing what is surely a sign of things to come,” said Hany Farid, an expert in digital forensics at the University of California, Berkeley, who reviewed the call recording and confirmed it is a relatively low-quality AI fake.

    The tapped and untapped value of AI tools to impostor fraud barons and amateurs alike has been on the radar for the past year. Banks, regulators, and various watchdogs would all like to be ready for the storm rather than trail behind an endless mess. I may be stealing the FTC's phrase "synthetic media" to describe what's now competing with .... uuuhhm .... 'organic expression'? We have already been living in the era of 'Variable Authenticity' (I'll be collecting royalties for that one.) for much longer than we've had the means to make disinformation as common as beach sand. We're removing the last barriers to entry, the need to be clever and industrious and experienced before graduating from petty crime to wholesale heist and disaster on a mass scale.

    The Family Crisis scammers and Bank Notice posers are early adopters, and I expect the Licensed Pro predators to join in soon. How long do we have before surveillance economics, data brokerage, and every wannabe hacker and phisher collide, and everyone with a pulse and a phone becomes a daily target of quite personalized threat actors?

    Somehow I imagine FCC Chair Jessica Rosenworcel sitting over a hot beverage the weekend after that bogus Biden call blitz, pondering what her agency can do to slow down the giant snowball without waiting for a blessing from a reality-challenged Congress. Her press release headline at month's end was characteristically blunt: "MAKE AI VOICE-GENERATED ROBOCALLS ILLEGAL"

    Her aim is a Declaratory Ruling to sweep voice-cloned audio under the TCPA umbrella which bars "artificial voices" from cold calls. Indeed, it is hard to conceive of a more artificial persona than one which eliminates human performance. I can't wait to see the chain-linked Mobius strip logic which autodialer-dependent businesses will invent to protest a simple classification effort. Some in the opposition remain on an Ibuprofen regimen from the bodily knots they'd formed while trying to convince us that soundboard operation ("speaking" only with canned audio files) should count as an exempt "live audio" call.

    She also revisited ongoing inquiry into leveraging AI to improve junk call pattern matching and maybe filter out the trash before it rings anyone's phone. Suddenly I'm thinking of Fred Saberhagen's man-slaughtering "Berzerker" series, and a set of stories from other invited authors. One of them has a space vessel crew discover a dormant Berzerker machine, and everyone studying it is terrified except for one woman, who realizes it was built to combat the enemy ..... an anti-Berzerker Berzerker.

    You realize what's the next frontier in phone fraud, yes? It's compressing all the functions of a call center into a single rack mount cabinet. Right now you still need human herds to react in real time to cold-called parties. One day the right software will replace all those dope fiend "qualifiers", never needing lunch breaks or maternity leave. At most your shop will employ one tech and a handful of sales closers. Otherwise it's 91% meat-free; the wet dream of Forrest Baker, the Reynolds clan, and a bunch of South Asian crime families.

    --------------------

    Voter Suppression Robocall Complaint to Election Law Unit - NH DoJ press release, 22 Jan 2024
    AG Office: Robocalls apparently faked by AI sound like Biden telling people not to vote - Concord Monitor, Concord NH, 22 Jan 2024
    New Hampshire voters report deepfake robocalls simulating President Biden's voice - WCVB-TV, Boston, 22 Jan 2024
    Bogus Biden robocall highlights concerns about expanding tool set for dirty political tricks - Inside with Jen Psaki segment, MSNBC, 22 Jan 2024
    Top House Democrat seeks DOJ probe of AI-generated Biden robocalls - Axios, 22 Jan 2024

    Thousands scammed by AI voices mimicking loved ones in emergencies - Ars Technica, 06 Mar 2023
    Chatbots, deepfakes, and voice clones: AI deception for sale - FTC business blog, 20 Mar 2023
    How AI Is Supercharging Financial Fraud -- and Making It Harder To Spot - Forbes, 18 Sep 2023
    'Grandparent' Scams Get More Sophisticated - FCC, 01 Feb 2024
  • +4
    Pudge replies to Resident47
    Thanks for posting...especially the links to AI scams.
  • +3
    Resident47
    Conspire In Style at The Game of Life

    By the Friday after the Sunday surprise calls to New Hampshire citizens, investigators at Pindrop Security had a smoking gun. The tool of choice was AI software produced by ElevenLabs, currently a hot investor property. "Officials at the company said they suspended the account of an unidentified user who created these calls", per Bloomberg and the New Hampshire Union Leader's curt reports.

    Forbes reporters got their hands dirty with the details of separate efforts to identify the software. A University of Buffalo professor and deepfake analyst took a run, submitting a call copy to the "speech classifier" managed by ElevenLabs. He said its in-house tool "is one of the first he checks when trying to determine an audio deepfake’s origins because the software is so commonly used". His guess was the impostors had used ElevenLabs, yet the same test conducted by Bloomberg returned "a 2% likelihood of being synthetic or created using ElevenLabs".

    It's known that ElevenLabs can detect its own raw output, but is deaf to the same file once it's been out in the wild and picking up stray noises and anomalies. This makes total sense to anyone who's handled audio recording across a distance. The Pindrop team had to gut the captured phone audio like a landed fish and compare it to their own stash of text-to-speech audio examples to peg ElevenLabs with confidence.

    The Wash. Post notes that "ElevenLabs has been criticized by AI experts for not having enough guardrails in place to ensure it isn’t weaponized by scammers looking to swindle voters, elderly people and others". Last year I had checked out the service for reasons I can't remember; it continues to promote itself as a low-barrier and cheaper alternative to all that mucking about with hiring fleshy actors for voice work.

    So within a week of the call blast we had the gun but not the shooter. How do you virtually dust this for fingerprints? Start with the credit card receipts, I guess. Remember when I said the guy in the call's end tag spoke in a Texan drawl? Today the FCC and the New Hampshire AG announced their findings and counterstrikes.

    Attorney General John Formella has sent his cease order and criminal probe intent to Life Corporation in Texas, likely source of the bogus Biden. (Follow your nose! It always knows!) A multistate civil case is warming up against Life and its carrier Lingo Telecom, headquartered in Michigan and tagged by the Traceback Consortium as the upstream provider. The FCC has meantime dropped the "K4 letter" anvil on Lingo, its now routine warning that a spamming carrier is at risk of losing access to the entire phone network. As with prior interventions, downstream carriers are now allowed to deny Lingo's auto-spam traffic. Lingo has two days to figure out if they'd like that condition made permanent.

    The Commission is not happy to confront repeat offenders with their lies about junk call mitigation compliance, including the fact that Lingo let its customer spoof the PAC lady's Caller ID. The FCC's nastygram reminded the carrier that Lingo has been considered a "gateway provider" of overseas spam since 2021. Following an Investigative Demand, cousin agency FTC stepped in by August 2022 and ....
        "issued Matrix Telecom, LLC, which was Lingo's prior corporate name, a Cease and Desist Demand that identified numerous illegal calls that the company was apparently transmitting".

    Post-name change, Lingo kept right on spewing "suspected illegal traffic", prompting the Anti-Robocall Multistate Litigation Task Force to holler at them last November. Lingo has managed to burn through ten other aliases, never a comforting sign: Americatel, BullsEyeComm, Clear Choice Communications, Excel Telecommunications, Impact Telecom, Lingo Communications, Matrix Business Technologies, Startec Global Communications, Trinsic Communications, and VarTec Telecom. Really if anyone needs a better name it's that Task Force. Tremble thee before "ARM-Littif"?

    Lingo's client Life Corp was cited by the FCC way back in 2003 "for delivering apparently illegal prerecorded and unsolicited advertisements to residential lines", in simpler days when copper landlines were still dominant. A reporter for TechCrunch, refusing to assume that little-regarded Life Corp has kept a shiny halo in all that time, critiqued its own promotion:
        "Life Corp.’s website isn’t much help; they have been “successfully creating solutions since 1987,” whatever that means. “WaWhen you are thinking, “I wonder if Life Corporation could do…?” The answer is probably YES.” [sic] “Please call use to discuss making a win-win arrangement for both of our companies.”

        It’s the kind of thrown-together gobbledygook that suggests the company might not really do anything at all, or whatever it does do can’t be said out loud or in public."


    The FCC transcript has filled what was lopped off from I'd found: "This coming Tuesday is the New Hampshire Presidential Preference Primary". A cooperating call blocker service figures "between 5,000 and 25,000 calls were made", relayed the New Hampshire AG's Election Law Unit today. I would not be surprised if that range becomes an undercount after the smoke clears.

    . . . . . . . . . .

    Biden audio deepfake spurs AI startup ElevenLabs — valued at $1.1 billion — to ban account: ‘We’re going to see a lot more of this’ - Forbes, 27 Jan 2024
    Voter Suppression AI Robocall Investigation Update - NH DoJ press release, 06 Feb 2024
    FCC Demands Entity Behind N.H. Robocalls Stop Illegal Effort - FCC press release and warning notices, 06 Feb 2024
    New Hampshire opens criminal probe into AI calls impersonating Biden - Washington Post, 06 Feb 2024
    AI-generated Biden calls came through shady telecom and Texan front ‘Life Corporation’ - TechCrunch, 06 Feb 2024
  • +3
    Resident47
    New Tool, Same Rule

    That no-brainer FCC Declaratory Ruling was quietly in the bag a week ago and announced yesterday, clarifying that fakie algorithmic voices are the same as "artificial or prerecorded" and therefore subject to the same TCPA barriers.

    Many news reports paint this as a jump shot reaction to the bogus Biden calls in New Hampshire. However, the FCC has been working on this issue since mid-November with one of their Notices of Inquiry, gathering opinions from stakeholders and attorneys general. Possibly the voter suppression stunt was at most creating urgency. Certainly it was a timely proof of concept.

    The vote to adopt was a rare five-zip. As is routine with published rulings, the three Democratic members each issued an attached statement explaining their votes. The red team kept quiet. Not much room in this small and easy rule adjustment for owning the Libs. By contrast, don't get Brendan Carr going on his Net Neutrality loathing; he won't shut up for a week.

    As a bonus, the Ruling thinks ahead to the same scenario I've posed, of some AI product turned loose in place of human headset jockeys:
        "... the TCPA does not allow for any carve out of technologies that purport to provide the equivalent of a live agent, thus preventing unscrupulous businesses from attempting to exploit any perceived ambiguity in our TCPA rules."

    The Ruling text echoed me again a couple paragraphs later, recalling the FCC's answer in December 2020 to an alarm seller and its hired call blaster who wanted to legislate their rule-breaking business models and silence lawsuits against them.
        "Our conclusion is also consistent with the Commission’s Soundboard Ruling, which confirmed that the TCPA applies to any telephone call that is initiated using an artificial or prerecorded voice message."

    Pressing buttons on a keyboard or mouse to generate canned audio, argued that Ruling, is not a "live agent" experience and therefore can't dodge TCPA liability. Besides which, those hidden agents are typically juggling three calls at once, using the soundboards not for meaningful discussion but to triage call recipients and rudely dump the non-performers. If you can't win an exemption with a mute wage slave in the room, don't even try it with only a circuit board in his chair, is how I read this pre-emptive strike.

    While we're on the "what-abouts", I seriously doubt the Feds would grant any more charity to the equivalent of real-time Auto-Tune for the English-deficient, discussed here a couple summers ago.

    The real sad part is that we needed this month's ruling at all, and by "we" I am glaring at a breed of corporate lawyers who think that obsessively scratching and peeling the seams of very sensible and much-needed laws is doing God's work. No one at the FCC, let us underline, imagines that career criminals with autodialers are gnawing their fingernails over some "by the way, this means you too" paper waving. The declaration does spare us months or years of catfighting over statutory prepositions and semicolons when "synthetic media" becomes a prominent material issue in the next big junk call beatdown.

    . . . . . . . . . .

    FCC Makes AI-Generated Voices in Robocalls Illegal - FCC Declaratory Ruling, 02 Feb 2024
    CGB Denies TCPA Petitions from NorthStar Alarm and Yodel Technologies - FCC "Soundboard rule" Declaratory Ruling, 18 Dec 2020
  • +4
    Resident47
    | 2 replies
    The Cost of Malarkey

    Today the FCC ratchets up the tension on the Bogus Biden bunch and names another gang member, "political consultant Steve Kramer". The FCC docs are still being published as I write, so it's less clear who he is. Maybe the drawling Texan heard in the canned call's end tag?

    The press release says only that "Kramer had engaged Voice Broadcasting Corp., which used the services of Life Corp. to transmit calls through voice service provider Lingo Telecom". Hammering the fact that the personal mobile number of a prominent party supporter was spoofed in the spam payload, the Feds want to send Kramer a six million dollar bill for the troublemaking.

    Meantime, the FCC proposes a $2-mil fine against Lingo Telecom for subverting the entire point of the STIR/SHAKEN plan. The naughty carrier "failed to follow “Know Your Customer” principles by applying the highest level attestation -- signifying trust in the caller ID information -- to apparently illegally spoofed calls without making any effort to verify the accuracy of the information."

    Given Lingo's history as a spam harbor, we may infer that devotion to accuracy would only wreck its business model. If the FCC has its way, the carrier will be first in history to be punished for its false seal of approval on fraudulent traffic. Hey, if you took home a jug of "freshly brewed iced tea" from the market and found yourself pouring used 5W motor oil into your acrylic tumbler, you'd be upset about that, too.

    today's press releases and commissioner remarks:
    FCC Proposes $6 Million Fine for Deepfake Robocalls Around NH Primary
    FCC Proposes First Robocall Fine for Caller ID Authentication Failures
  • +3
    B-Edwards replies to Resident47
    | 1 reply
    Since we are in an election year - I am sure you remember Alan Abel the prankster. Omar's School of Begging and such. He did political pranks as well. Or Donald Segretti who ran the dirty tricks against the Dems in '72. Think of the damage that those characters could have done with AI.
  • +2
    Pudge replies to B-Edwards
    Wow...I forgot about those "people."  Thanks B.

Reply to topic