I am considering making a website where people upload videos and they are paid an amount of money that depends on how long others watched their videos (or number of views).
The main hurdle I see is combatting fake views. For example, a video creator uploading a video, then playing it hundreds of times on repeat in multiple browsers to increase their count.
Are there any open-source libraries or does anyone have suggestions for how I could write something myself which could achieve what I’m talking about?
For example, the video creator turning on a VPN to hide their IP and playing their own video over and over, or using a bot to play it in massive quantities.
Basically, similar to what Youtube must do to prevent content creators from earning ad money through fake views.
So how do you identify a user who’s playing their own video behind a VPN on Incognito? (You can’t. That’s the point of a VPN and Incognito.)
You could go find lists of VPN servers, and ignore all views from those IP’s, but that would be preventing legitimate views also. So there’s a tradeoff there.
You could count multiple rapid views from a single IP as 1 view; but whats the definition of rapid? If I like a video and watch it again a couple hours later, is that 2 views or 1?
The point i’m making is theres no easy answer to this. It’s not as simple as “ban the fake views”. You have to be able to measurably define your border of fake and real.
Do YouTube have countermeasures in place? Absolutely. What are they? If I knew that and I was a bad actor, I could get around them. So YouTube dont give out their secret sauce.
I know it can’t be perfect, so I want to start with the obvious ones (extremely rapid repeat views from same IP, like no breaks in between views for hours) and adjust as far as I feel comfortable. I was hoping this was a problem at least partially solved through through someone’s pre-made library using analytics wizardry. Maybe the main thing will have to just be rapid repetitive views.
And maybe as I view the logs I’ll see other patterns I can try to combat, such as the same IP viewing a video at the exact same time, to the second, every single day.
I can’t say I’m aware of any predefined libraries for this sort of thing (there aren’t a whole lot of CPM video services out there, as you can imagine); that isn’t to say that something doesn’t exist, as I’ve not had cause to dive deeply into the subject myself. I would imagine most solutions are paid services (something like a tighter defined Cloudflare or such DDoS protection type solutions).
Doing it yourself is going to get deep and quick. In the short term, if you’re looking at JIT processing, you could theoretically track incoming IP’s to video pages, and lookup the last time the IP looked at a video; if less than <insert some threshold here>, ignore the view, else count it.