Stories
Slash Boxes
Comments

SoylentNews is people

posted by Fnord666 on Wednesday September 06 2017, @12:45AM   Printer-friendly
from the another-job-lost-to-automation dept.

Spotted on HackerNews is a link to a paper on Automated Crowdturfing Attacks and Defenses in Online Review Systems:

Malicious crowdsourcing forums are gaining traction as sources of spreading misinformation online, but are limited by the costs of hiring and managing human workers. In this paper, we identify a new class of attacks that leverage deep learning language models (Recurrent Neural Networks or RNNs) to automate the generation of fake online reviews for products and services. Not only are these attacks cheap and therefore more scalable, but they can control rate of content output to eliminate the signature burstiness that makes crowdsourced campaigns easy to detect.

The paper, available from the arXiv link, contains the details of the attack which the paper notes "are largely indistinguishable from real reviews to human readers", and suggests defensive mechanisms based on "statistically detectable variations in the character-level distribution of machine-generated reviews".


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Wednesday September 06 2017, @03:06AM (2 children)

    by Anonymous Coward on Wednesday September 06 2017, @03:06AM (#564022)

    Negative "verified purchase" reviews can still be removed by the seller with no recourse by the reviewer except to resubmit another review. They call it "redacted".

  • (Score: 2, Troll) by aristarchus on Wednesday September 06 2017, @08:51AM

    by aristarchus (2645) on Wednesday September 06 2017, @08:51AM (#564084) Journal

    All this sounds strangely familiar. . . .

  • (Score: 0) by Anonymous Coward on Wednesday September 06 2017, @09:52AM

    by Anonymous Coward on Wednesday September 06 2017, @09:52AM (#564094)

    In that case, you know the review system is useless - goto yelp.

    If sellers can "redact" negative reviews on demand, seems to follow that people should also be able to "redact" negative items on their credit reports as well.

    To me, the negative reviews say far more than the positive reviews do. To get around this, I have seen people ranking something 5 stars, just to make sure its not redacted, then in the text they spill the beans on their experience..... after reading between the lines, its painfully obvious the buyer was super-pissed.

    The thing I fear is the machine reviews may actually imitate a pleased customer.

    Right now, its pretty easy to detect reviews made by "internet reputation businesses", as they read just like an ad. You know you are being manipulated and already know once you see this, you know you are dealing with a business which considers deceit as a business method, and re-evaluate your purchase decision knowing the company uses manipulative techniques to find a sucker.... you know, the same kind of respect loss one has for a business after seeing a TV ad which used deceptive language that you caught onto before making the call.

    I see so many TV ads that are so deceptive I wonder why anyone calls 'em. You know things like

          "We will send you another one FREE ( just pay separate fee )",

          "with autodelivery" ( now they have your credit card number and won't stop sending junk ),

          "when bundled with" ( gotta buy some way overpriced thing you don't want to get the thing you do want ),

          another variant... they will yammer on about phone, TV, internet, then right at the very end, recite the price, "each".
    gotta buy 'em all to get any of 'em.

          "order now! Try for 90days! Just $19.95!" ( The trial is $19.95, once you get it, the onus is on you to get it back to them. The thing costs $1000! )

          "no cost to order!" ( useless until "activated", which will cost you an amount which they are reluctant to say on the air.)

        and my all time favorite:

    Diet Pills. We sold 25 million bottles! We took two groups of people, one group asked not to change diet or exercise, Our group lost 400% more weight!

    ( Yeh, the group that did not change diet or exercise lost 0 pounds. 4 * 0 pounds is still 0 pounds. ) They found 12.5 million ( 2 bottles per order ) people who would send them at least $30. That's almost $400 million dollars for pills they implied do nothing. My guess is most of the proceeds go right back into TV advertising to rake in the next round of people who will send them $30 because some guy said "lose 400% more weight" through a microphone, without checking the context. I would find it interesting to know how many people have to send them $30 to justify running the ad? Its obvious a lot of people are sending money and likely getting themselves snarled up in autodelivery schemes, as those ads seem to run nearly every commercial break. Same as those senior life insurance ads. If those seniors want to "not leave their loved ones with their debt", open up a savings account. Do the damn math! No wonder those ads run all the time. Real moneymakers if you can find people willing to pay a buck for a 50 cent return!