A source for this picture? | Hacker News Reader

A source for this picture?

twitter.com

175 pointslfpa282 comments

Just to head off a lot of confusion stirring on the net about this: The person who posted the picture is claiming this woman is doing app ratings, not some outsourced App Store submitted app approval process.

Ah interesting, as in: she's the person who rates your app if you "buy" some ratings?

Yes. It's funny because everyone knows this has been going on forever, but I guess once they put a picture to it, everyone gets up in arms.

And I bet if you wanted she could rate your competitor's apps as well

I'm pretty sure a friend's app got a bad rating when he declined to sign up for a service like this.

How did that go down? You would think it would be easier to just not talk to someone who solicits this.

My understanding is that he got dozens of offers to help promote his app. Some were blatant buy-a-download - just to gain app store rank - some were less blatant, others were more traditional buy-a-targeted-clickthrough. It can be difficult to tell at first glance.

Thanks, that does indeed help head off confusion. I was thinking age ratings, but those are self-reported, and any checking on Apple's end is just part of the overall review process.

But when you look closely, the screens show different images. Could it be that she is farming some in-game accounts to higher levels? Or dating?

Or it could be ad click fraud

I'm surprised the crowd here is surprised... you can also hire Chinese to solve catchpas, etc. and it's on the cheap side of things.

plenty of captchas solved by presenting them to pron sites visitors

I definitely wasn't surprised, as I've known this principle has been going on for some time. But I, for some reason, thought this could be done somehow with a single device, or something. I never really knew how it happened, just knew that it happened.

So this picture answered that question for me (though I'm sure this isn't the only way it can or is being executed).

She could possibly make more money touring with a sound mixer and synthesizer apps on all those. I'd go see some live complex drone music, I'd help her set it up.

It's quite probable if someone is manually entering ratings on hundreds of phones, that the owner of the phones is not the same person doing the labor.

Is there a reason that you think she, in particular, would make money using a synth app on those phones, versus anyone else?

If she doesn't own them she should consider asking her boss if she can open for Melt Banana sometime.

What would happen if you used each phone to call another phone on the rack and then put them all on speaker phone? The feedback would be fascinating.

What if you created a FaceTime array in which each phone connected to an audience member's phone? You could then place tone emitters throughout the venue and the physical placement of each participant would shape a chord assembled on stage.

So much room for creativity beyond the creative grey market businessman's greed. That was the point of my post.

There's room for how many app store ratings companies/employees? With a reasonably stable income, too. There's room for how many iPhone feedback musicians?

In fact, what you're suggesting may have already happened. Boss took a set of iPhones, then ran a band. And then bought more iPhones to grow his ratings business. So the image you see would be the same, regardless.

This looks something quite easy to detect and ignore if apple was serious about ratings

There's no reason to believe they are serious about them. Search and ratings in the App Store are both awful.

When people have a company based around abusing your systems, it's never "quite easy" to get around them.

How would you detect it? Any detection algorithm that springs to mind seems easily defeatable. These devices are most likely all on individual pay-as-you-go cellular cards, each with their own iTunes account. Each review is probably either only a star rating, or for text-based reviews a randomly generated unique paragraph.

You could detect similar texts but that's just an arms race against new corpora being added to their generator.

Maybe detect the same app getting a lot of similar ratings in a period of time? But then the farm could just randomize the input list of apps among the farmers and stretch out the time period to make it look like more natural traffic.

Thoughts?

I think that those kinds of things are necessary to fight spam. Gmail isn't great at detecting spam because there's some invincible algorithm invented 10 years ago that catches everything and hasn't changed. To block spam, Gmail is always adapting it's spam filtering bases on behavior of spammers.

I believe one would probably have to do the same thing to moderate app store rating.

multiple high ratings for the same app coming in over a short period of time via the same external IP or GPS location?

You could avoid that by 1. having many apps up at once instead of just one. 2. Probably isn't hard to use a proxy for each device.

Those iPads could easily be routed through different proxies.

The IPs would be different with different SIM card vendors. The GPS location can be easily disabled on the device.

Reviews are tied to app store accounts, which are tied to credit cards. Reviews also reset with each version of the app. You could detect the same accounts being used to review each version of the app, discount ratings from newly created accounts, discount ratings from accounts with reused credit cards, etc. It seems like an easier way to game the system would be to pay anyone with an iPhone $5 to download and rate your app.

There's lots you can do to detect "false inputs" like this. How hard this behavior is policed is based on how badly Apple cares about this problem.

All evidence points to Apple not really caring about abuse of this variety, so the obfuscation needed to pass any filters are pretty basic.

The reviews themselves are near worthless, but in bulk they start to gain value.

It would probably be better to cluster reviewers by apps they've reviewed, and to try to find patterns that way.

You can have an app store account without a credit card. I don't think it was always possible, but it is now. Also, the accounts don't have to be newly created. This operation could be quite long-lived.

You could have some automated system which loads many different apps to be rated, so that the person in the picture isn't rating the same app over and over again in a short period of time.

Amazing illustration of a Sybil attack

http://en.wikipedia.org/wiki/Sybil_attack - didn't know that one: "The Sybil attack in computer security is an attack wherein a reputation system is subverted by forging identities in peer-to-peer networks. It is named after the subject of the book Sybil, a case study of a woman diagnosed with dissociative identity disorder. [...]"

I'm impressed the first reply came from someone from WSJ looking to make an article I presume.

It doesn't look like there is any EXIF data in the image, perhaps because it's been opened and saved so many times. If we had the original image, that might shed some light as to where/when it was taken.

Twitter strips EXIF. Most big sites do.

Tineye also shows no hits on that image.

I was half-hoping that this image was somewhere...

why do they strip the EXIF?

i imagine it is to protect the identify and location of people who post images and have no idea about exif/how to strip it.

Also saves us who do know from the hassle of stripping it out on our phones before posting ourselves.

Fiverr's headquarters exposed.

Gaming the system will always be apart of the game.

Seems like this could be about app store ratings, but it seems more likely that this is about manual testing. Wake me up when there's some more (credible) information.

That seems like an awful lot of iPads for one person if she's just doing manual testing.

Well, they might be testing many apps at once. A lot of bugs are the "well it was running a long time and it crashed" type that you'd outsource.

All the devices look the same which would be strange for manual testing. However, agreed that to assume it's app store rating is a logical leap.

Well, the source of the image claims that it's about App Store ratings. It's not like we're making an intuitive leap here.

The submitter is asking for an original source, not for the very source that he or she submitted. The person who tweeted it doesn't know any more than we do.

word.

If the paid-for app ratings don't do it like this, I would be surprised.

Honestly I would. I figured it'd be reverse engineered and scripted rather than actual physical devices

The technical term is "growth hacking"

lol

In that pic there are roughly 18*6=108 devices per panel. On the other side of the girl it seem there's another panel which could have another ~108 devices. That's 216 good reviews/votes coming from a single asian lady... notbad.jpg

I wonder if it is possible to reverse-engineer the whole iTunes connection and review process, and send fake reviews automatically from a server, or even from a botnet.

That's why I feel lucky that my very first attempt to make a food rating app for China market failed at product building stage almost 3 years ago.

As replied the OP, it's an anonymous friend of him.

He says in his very next tweet that he doesn't know if it's real:

https://twitter.com/simonpang/status/562201979603021824

There’s lot of pay-to-review service on the net but most ppl never seen how they work. Dunno if it’s real but still give us some insight.

There's a pretty big "if" in that sentence.

Yup! So important I repeated it.

Can't you do this more efficiently opening several instances of an emulator on a powerful computer?

Not for iOS

You wouldn't need iOS, you could do it perfectly fine in iTunes on Windows or OS X.

1) Setup VM with OSX, 2) Sign-in iTunes account created with a gift card, 3) Download/purchase app in iTunes (OSX), 4) Rate app in iTunes (OSX).

Step 3 might require a physical device, but 1, 2, and 4 do not.

Wouldn't it be easier to just use Mechanical Turk for this?

wouldn't they need iphones?

And everybody shits on google for trying to tie personalization to everything.

Do you want to sort your apps by the the rank and reviews of chinese laborors? Or do you want it to prioritize ratings by people you actually know?

Personalization is really the only way to nuke the arms race that is fraud/spam detection and crowd-sourced data.

There is no perfect answer. Just more abstract games to play.

Would I like more privacy... or more accurate app reviews... Hmmmm...

Na they would just create 1000's of personalities. Maybe by IP Address and time with personalities would work or even better stop ranking by it at all.

I would say having a group of trusted reviewers would be the best way to handle that. Nothing is a good substitute for trusted content curators.

> I would say having a group of trusted reviewers would be the best way to handle that. Nothing is a good substitute for trusted content curators.

The point they are making is that with Google, your friends app reviews matter more than random peoples.

But I also follow my non-friends as well, I also might trust one friend more then another for their opinions.

That's a false dichotomy. The choice is not between tie your identity to everything or Chinese rating farms (nice work invoking the scary Chinese other as some specter that is ruining our app processes by the way - remember when the Japanese were ruining everything in the 80s?). There is enormous room for improvement in app rating procedures and process. Saying the only choice we have is what you consider the lesser or two evils based on the current state of the current system is almost laughably short sighted.

I'll gladly take privacy over useless albeit "authentic" app store rankings.

I'm sorry to break it to you... http://thenextweb.com/facebook/2014/02/03/facebook-estimates...

How she operates that many phones?

I assume every phone is a different user, and she just goes around rating apps. I am not sure if you could automate that process, but then why would the phones be arrenged in such a way that one person can quickly click on every one of them, instead of a dark room.

Probably could automate it, maybe even by using a Mac and multiple instances of the iOS simulator, but that would require a lot of expensive development, so it's likelt much cheaper to buy a bunch of devices and hire super cheap labor.

No App Store on the simulator.

I don't believe you can rate an app on the app store from the emulator itself. You need a device.

Or it's something else. Maybe a dating app?!

You could build a hardware rig that's programmed to hit the 5-star rating area, and then "submit" - but you'd also need to find a way to launch the app store to the correct app every time, and then debug the thing, and then still pay someone to stand around and watch in case it gets into trouble anyway.

Open CV could easily help in that area. Servo controlled stylus for input, OpenCV based watchdog against pre-programmed motions. Maybe mechanical turk or something to generate the reviews.

Perhaps you could fight it by doing statistical analysis on review patterns. If the device is only ever giving great and/or crappy ratings and the same group of devices rates the same group of apps the same way, you could detect this.

Edit: Of course, they could fight detection by having a queue of apps to review and an array of devices. Randomize the matrix and periodically retire a device to prevent it from gaining too much history. Since it's running in a machine it would likely be in pristine condition and capable of being resold as new. Heck, they could buy one, use it for a period of time then return it.

I guess someone will eventually build one and start selling it to those rating-fraud companies.

Just yell HEY SIRI :)

Reply on news.ycombinator.com