Artificial Intelligence in the Mail.ru service

Good day, dear Habrasoobshchestvo!
I have long and tediously thought about what will be my first post here.

And today I decided to talk about artificial intelligence and stolen (stolen) accounts Or rather, how the first can come to the rescue and fight the second.

All names and addresses mentioned in this article are fictitious. Any coincidence with real people by chance.
')

Prehistory

It just so happens that I am a user of Mail.Ru for many, many years.
Being engaged in offers of friendship, I was faced with the situation of outright porn spam, both immediate and deferred.
(This is about accounts that advertise all kinds of services that I do not use in particular demand.)

Solution (currently manual):

1. Check accounts by the list of common friends. If there are friends on this list who enjoy my trust (they don’t keep spammers on the list of friends), then this address can be entered into your friendship list.
2. Check accounts for synonymous usernames in a postal address, for example: Marina Petrova (marishka @ mail domain, m.petrova@mail domain, mary @ mail domain and their combinations)

Since laziness is the engine of progress (I think the programmer colleagues will agree with me), it was decided to make a tentative attempt to cry out the development of an artificial intelligence algorithm capable of solving this problem with minimal human involvement.

I invite you to discuss.
UPD: All the necessary technologies on Mile already have.