Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

nytimes.com

Post reply on HN
Print view View original

31–40 of 124 posts

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#31

Post by rjtavares »

jmcqk6 wrote: post #28
menacingly wrote: post #5

I get that facebook my have complicated motives here, but I don't know that "easy to spot" is a good yardstick. First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. Second, and perhaps most importantly, I don't think simple hunches about fake accounts are really actionable here. Having an account that "originated in the Middle East" that now presents…

>First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. This is an interesting argument to me. It's basically conceding a shittier solution, because we "can't scale" the effective one. As we talk more and more about automation, I think this is similar to what we've seen in meat space with things like tools. I've heard so many people complain about chine…

The way developed countries handle that is through regulation (specifically, fines). That changes the cost equation.

Eg. Germany passed a law that fines Facebook if they don't comply with German hate speech laws.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#32

Post by johansch »

marcinzm wrote: post #16
menacingly wrote: post #5

I get that facebook my have complicated motives here, but I don't know that "easy to spot" is a good yardstick. First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. Second, and perhaps most importantly, I don't think simple hunches about fake accounts are really actionable here. Having an account that "originated in the Middle East" that now presents…

The issue is also one of false-positives. Even if you could spot fake accounts 100% of the time if you also flagged 1% of legitimate accounts as fake by that approach then it's non-viable. Remember that at Facebooks scale a 1% false-positive rate means 20 MILLION accounts. That's too much to even try sending for manual review.

A manual review would cost what, a dollar at most?

That's about 1/2000th of FB's yearly revenue. They should be able to handle that.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#33

Post by pbhjpbhj »

marcinzm wrote: post #16
menacingly wrote: post #5

I get that facebook my have complicated motives here, but I don't know that "easy to spot" is a good yardstick. First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. Second, and perhaps most importantly, I don't think simple hunches about fake accounts are really actionable here. Having an account that "originated in the Middle East" that now presents…

The issue is also one of false-positives. Even if you could spot fake accounts 100% of the time if you also flagged 1% of legitimate accounts as fake by that approach then it's non-viable. Remember that at Facebooks scale a 1% false-positive rate means 20 MILLION accounts. That's too much to even try sending for manual review.

Trust networks seem to work here: for example I've reported accounts like "Blacksmith Arms" (not a real example) for being not a real person. Blacksmith Arms is a common pub name, it's a business account masquerading as a user which is (or was) against Facebook rules.

Facebook responded on a couple of such reports but the accounts weren't altered.

If my account had sufficient standing then the action could have been automated. My accounts actions could then be meta-moderated by other accounts that are more trusted.

You make decisions using reports from trusted accounts, and verify them with multiple more-trusted accounts. Verified actions improve an accounts trust level.

Surely Facebook already use such a system? I imagine Google Maps are using such a trust network too?

Use the scale of the system to manage itself.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#34

Post by nmstoker »

This needs care, but I don't see why they can't just rate limit key activity from high risk accounts that are reported, then let the person carry on. If they're genuine then things like friend requests will get accepted but if they're fake, those will get a poor response and seal their fate.

Simple: earn trust through good behaviour. And it's not too heavy handed for the false positives.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#35

Post by wpietri »

jmcqk6 wrote: post #28
menacingly wrote: post #5

I get that facebook my have complicated motives here, but I don't know that "easy to spot" is a good yardstick. First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. Second, and perhaps most importantly, I don't think simple hunches about fake accounts are really actionable here. Having an account that "originated in the Middle East" that now presents…

>First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. This is an interesting argument to me. It's basically conceding a shittier solution, because we "can't scale" the effective one. As we talk more and more about automation, I think this is similar to what we've seen in meat space with things like tools. I've heard so many people complain about chine…

Yes. And the way it's often argued, it treats the status quo as some sort of natural given, like the weather. The line of argument seems to be that if Facebook and Twitter can't run safe platforms while maintaining their current revenue models and profit margins, well, we'll just have to live with that lack of safety.

I find this line of argument odd when it's about actual natural givens, like dangerous weather. I'm in a house right now, so it's not like technology to limit the effects of dangerous weather is unheard of. And a bit odder when it's about human inventions that have been around for generations, like guns and cars. But when it's used for something basically new in the world? I find it baffling.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#36

Post by menacingly »

jmcqk6 wrote: post #28
menacingly wrote: post #5

I get that facebook my have complicated motives here, but I don't know that "easy to spot" is a good yardstick. First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. Second, and perhaps most importantly, I don't think simple hunches about fake accounts are really actionable here. Having an account that "originated in the Middle East" that now presents…

>First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. This is an interesting argument to me. It's basically conceding a shittier solution, because we "can't scale" the effective one. As we talk more and more about automation, I think this is similar to what we've seen in meat space with things like tools. I've heard so many people complain about chine…

This is less like a nut that needs turned and more like getting the courts to consistently put "bad people" in jail or define "obscene". Just because it has to do with software does not mean it is a problem software can solve.

Even if it is a human doing it, it still has to scale to many humans, and there are plenty of examples of things we have never manged to get right due to scale well outside of technological limitations.

Even if you could, my example from the article was one human, with one example case, and as another human I didn't see it as a cut-and-dry bot. Try doing that with 20,000. It's not that it would be "costly", it's that it would be an unavoidable train wreck

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#38

Post by ghostbrainalpha »

olivermarks wrote: post #3

Facebook's login page encourages you to set up a new account ('always free') despite their knowing you have an account. Makes me wonder how many sock puppets and variants there are out there. Regarding 'Russia' surely we should be saying 'Eastern European Gangsters?' rather than a nation state...

I wonder how much of the problem would be eliminated if Facebook accounts had a 1 time $5.00 set up fee?

I'd definitely be willing to pay a 1 time fee for a 90%+ reduction in fake accounts.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#39

Post by mikestew »

jmcqk6 wrote: post #28
menacingly wrote: post #5

I get that facebook my have complicated motives here, but I don't know that "easy to spot" is a good yardstick. First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. Second, and perhaps most importantly, I don't think simple hunches about fake accounts are really actionable here. Having an account that "originated in the Middle East" that now presents…

>First, there are the countless instances where something is trivial for a single person to suss out that doesn't scale. This is an interesting argument to me. It's basically conceding a shittier solution, because we "can't scale" the effective one. As we talk more and more about automation, I think this is similar to what we've seen in meat space with things like tools. I've heard so many people complain about chine…

Things mass produced cheaply are lesser quality, but we're willing to make the trade off as a society because we get things for cheaper.

Problem is, from what I'm seeing, is that the market bifurcates into "shitty, but cheap" and "OMG, the price, but quality". I want the middle way: "reasonably priced, 'good enough' quality, and won't break the bank". Tools are a prime example. You had Snap-On/Matco/etc., Sears Craftsman, and "stamped steel sockets from China". Snap-On for the pros (and worth every dime if you're a pro), Craftsman for the guy with the '57 Chevy he works on regularly, and then the cheap shit for the person that needs to turn a 13mm nut once a year.

Now we "get things for cheaper", but those things (in the case of tools) are not something I'm going to let anywhere near a fastener I care to ever reuse. Craftsman is gone, so my option is to chase a Snap-On truck around town [0]. I don't need Snap-On quality, it's like buying a $100K CNC machine to turn a table leg when a lathe will do. I just need a decent socket that won't round the fastener.

Seems like the middle fell out of appliances, too. Only in this case, it looks to me like the high-end is shit, too. Every single person I've known that's had a Viking stove or SubZero fridge has had it worked on at least once. I think our Kenmore stove/fridge and LG washer/dryer are disappointingly not built for longevity, but at least nothing has broken after five years. But that washer isn't going to be washing clothes twenty years from now like old Kenmores.

About the only thing I can think of where the quality goes up and the prices down are cars. Man, it is amazing how long a car can go while doing practically nothing for maintenance. And they are much more complex systems, and live in much harsher conditions, than a washer.

To tie this back to the topic at hand, "we can't be bothered to take the time to make it nice" is a fine excuse if only they'd use that as the actual excuse. "Doesn't scale" just means "too much trouble to make it nice, so you'll have to settle with half-assed". If "that's just the way it is", then I don't want "it".

[0] When I bought Snap-On, it was from a truck that drove from auto shop to auto shop, selling tools out of the truck. They had no retail presence; IOW, you couldn't go down to the Snap-On store, it all came off the truck. Not a practical option for the home mechanic, though I've heard they sell online now.

Re: Facebook Says It’s Policing Fake Accounts, but They’re Still Easy to Spot

#40

Post by Stranger43 »

rosser wrote: post #30
Stranger43 wrote: post #21

Earlier quoted context omitted.

Except of cause that's also how very real grassroot movement's look especially those not made up of American college students, i.e. you have a fairly high chance of chance of you accidentally targeting a few thousand genuine users. Remembering that almost all of facebooks growth potential is outside of the US. Facebook really cannot afford to just kick off people based on not being us college students. The problem is…

> ...you have a fairly high chance of chance of you accidentally targeting a few thousand genuine users. Which is why I said "further scrutiny", and not robo-ban. The tone of these posts is very much, and very consistently the whole "sow dissension" and "poke at wedge issues" shtick that is demonstrably part of how the Russians have been meddling in our shit — and which Facebook, perhaps knowingly, profited from. (Un…

The college student problem is btw a very real concern in social science i.e. way to many studies have been made using local college students as study subjects, which is useless in a world where the average person speaks bad English and think the US is the biggest threat to world peace(however wrong that view might be). I.E. you cannot and should not build your algorithm about what is normal for your local but have to look at data that might not be available to people other then Facebook.

Foreign influence is a reality in every single election, worldwide, to the point where nobody have even accused Russia of doing something that the US would be comfortable making illegal when it comes to US based organizations interacting with foreign nations.

The problem is that the entire debate starts from an obsolete(fake past) view on how interstate relations work in the modern world where the diplomatic corp plays the only major role, where as the reality is that with the modern Internet traditional diplomacy have given way to a much more direct people to people interaction meaning that you cannot just assume that a post being foreign and exposing a viewpoint that's not organic to the beltway are the product of state controlled a botnet, it might be a foreign movement deciding to engage in modern direct diplomacy.

And Facebook cannot grow while remaining an US centric platform so the commercial drive for facebook is to become the facilitator of direct people to people diplomacy rather then a defender of the old diplomatic model as some American politicians obviously want them to be.

And if you cannot use content or "national style" as a guide to spot bots you have to depend almost entirely of metadata about the IP or client involved especially as false positives based on content analysis have a potential for market backlash.

Post reply on HN