Aug 10, 2018, 16:51 pm
(Aug 08, 2018, 04:44 am)jerrysmatrix Wrote:This is probably the only way, I agree. The algorithm thing would be a great working solution I think. The way the spammers behave is practically the exact same pattern over time, so this would be great if we can do it. I use linux so .exe and viruses don't really affect me.(Jul 30, 2018, 19:32 pm)politux Wrote: Then they just open 10 accounts and upload 15 torrents with each account....
I'm not sure what systems are currently in place right now, but there's always a lot that can be done to help.
You could base it by IP address, but that would cause issues elsewhere.
You could limit 10 torrents per account, and limit the new accounts by IP, and raise flags based on that.
You could gather statistics and data on known spammers and fakers, and manually build rules based on that or feed it into a machine learning program.
You could implement a voting/rating system that hides torrents or warns people, and weigh the votes based on seniority and trustworthiness so that it doesn't get abused.
You could show a warning on all posts that violates these rules, and remove the warnings as they get validated.
When all else fails, you can make systems to bring suspicious torrents to the front that require attention or validation.
With all of that said, I'm quite sure someone could fairly easily come up with an algorithm to effectively detect spam posts, because all spammers/scammers tend to follow patterns. If it is an algorithm with too many false positives, it could in the least make it easier to bring to a moderator's attention or warn others. The trick is to find the right patterns to filter out.