You are viewing a single comment's thread from:
RE: How spam filters save us from intrusive ads (featuring @phenom as author)
I expected to read about Bayes filtration...
Bayes' nets (filters) makes a revolution in email spam detecting around two decades ago.
But they are useless for messenger spam -- messages usually are too small.
"Hello!" -- is the spam message or it is not?
You start to message and when it is enough information for filter to classify spam / not-spam, you understand it by yourself.
But the true challenge for spam is the human-bots-spam, men are spammers, globalization makes human-spam cheap.
You can buy 1000 anticaptcha made buy indians for $3.
You're absolutely correct. Bayes filters made a great impact on the development of spam detection. I didn't drill down into technical details here and didn't tell about Bayesian filters as want to write a separate post devoted to them (especially to Kalman and Particle filter). Thanks for mentioning filtering of human-bots-spam. It's indeed a big challenge as an ad in messengers sometimes is utterly annoying. I think in this case messengers also use the probabilistic approach, e.g. if this contact is not from your contact list then probability that it's a spam is much higher.