5 ms·
One thing that doesn't make sense is surely a service like this greatly benefits from the network effects of being able to see activity across multiple sites, s
by snowwolf 8y ago
One thing that doesn't make sense is surely a service like this greatly benefits from the network effects of being able to see activity across multiple sites, so if Twitter wanted to continue using the technology, wouldn't they have been better off leaving it running?
- make3 8y agoI'm not convinced of your premise, spam is adversarial & spam detection is just pattern recognition. Twitter likely doesn't want to allow people to be able to test against their spam shields by spamming other people
- snowwolf 8y agoAs I understood it they were using machine learning models to identify attacks. Abuse and fraud strategies are constantly evolving, so the benefit of having many clients is that once you see a new attack strategy at one, you can instantly protect all your clients from that threat. In your own example "Twitter likely doesn't want to allow people to be able to test against their spam shields by spamming other people" Yes, the attacker may find a new pattern that can bypass the shield, but as soon as that pattern is added to the machine learning recognition, they are then protected against it on their own system. "spam detection is just pattern recognition" The more patterns you can analyse, the better your recognition?
- dredmorbius 8y agoTwitter may have sufficient scale that the additional data aren't meaningful. Friend worked on a large, multi-institutional risk-modelling project ... until the largest institution decided it had enough data on its own, and didn't need competitors who were competitive in this regard, and pulled the plug. Law of large numbers (statisttics) is powerful. Even for complex (multvariate) models. It sounds as if there may well be other considerations at play, but I'm answering your question specifically.