Improve antispam

[Chiiruno]

There are several patterns a spammer can spam meguca, we can block some of them.

• If the post contains a name/date/post number multiple times in the body, it is likely spam.
• If the post contains multiple (like, 10) broken links, it is likely spam.
• Perhaps a way to detect complete gibberish posts that exceed ~500 characters?

We can just void these types of posts, increment a counter, and if the counter reaches a maximum of X, ban the poster.

[bakape]

Need to consider tradeoffs for performance, storage capacity and how to do indexing and lookups.

[Chiiruno]

Checking broken links will be too much of a performance penalty per-post, scratch that one. Checking complete gibberish is pointless, since gibberishposting is fun. Only viable anti-spam I see here is detecting mulitple dates, or variations of relative time.

[bakape]

Checking broken links will be too much of a performance penalty

All links are already checked for being broken on post close.

[Chiiruno]

Okay, I'll have to consider that.

[Chiiruno]

Consider excessive command use over 2 or more posts to be spam and auto ban.

[bakape]

No need. Comads will all be cheap with the removal of roulette.

On Fri, 8 Nov 2019 at 07:43, チルノ notifications@github.com wrote:

Consider excessive command use over 2 or more posts to be spam and auto ban.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/bakape/meguca/issues/910?email_source=notifications&email_token=AB347MEFLGJ6TMXXOCXUR5DQST4A7A5CNFSM4F7WJID2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDOZQHY#issuecomment-551393311, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB347MBIVCVSQTFSV4DYOGLQST4A7ANCNFSM4F7WJIDQ .