On those simple ideas to fix it: I don't think it's simple. Once you do simple heuristics, the other side will start doing it just a little bit more sophisticated, to get around the simple heuristics. So, then you improve the heuristics to catch the spammer. And again, the spammers get around that improved tests as well. And so on.
In the end, you end up with similar spam filter methods as we also have for mails and probably as other social networks have as well. But this is far from simple. I don't think having a huge number of hand-crafted heuristics is really a good solution. I think it should be machine learning model which you train and it does it all automatically without too much false positives (and also not too much false negatives).
In the end, you end up with similar spam filter methods as we also have for mails and probably as other social networks have as well. But this is far from simple. I don't think having a huge number of hand-crafted heuristics is really a good solution. I think it should be machine learning model which you train and it does it all automatically without too much false positives (and also not too much false negatives).