HackerTrans
TopNewTrendsCommentsPastAskShowJobs

white_viel

no profile record

Submissions

GPT Bot Ignoring Robots.txt on my cloudflare worker

3 points·by white_viel·hace 5 meses·9 comments

Plex Update: Notice of a potential security incident

34 points·by white_viel·hace 10 meses·12 comments

comments

white_viel
·hace 5 meses·discuss
but i am impressed that gpt was able to decipher the worke.dev url on its own and access it.
white_viel
·hace 5 meses·discuss
>> interesting. workers.dev domains can be a liability sometimes -- if you've mapped the worker to a real zone, then you probably don't want the workers.dev zone anymore.

maybe that is a good idea, prompt a user to disable the worker.dev domain. however, having the worker.dev domain open for me is more of a backup way to access the worker as the mapped domain is more for a hobby project.

> Interesting. workers.dev domains can be a liability sometimes

what about allowing the specific worker.dev to serve a robots.txt on its own that can be used to disable the AI bots?

i.e https://XXXXXXXXXXX.username.workers.dev/robots.txt should be allowed to be configure on a worker level. not sure how that affects the design of the infra, but it would be a good idea
white_viel
·hace 5 meses·discuss
Finally managed to kill the traffic: 1. renamed the worker - the bot was bypassing the route on the domain and using the worker.dev domain 2. add WAF rules to block gptbot in user agent 3. serve zip bomb on request from GPT.

interesting that GPT was accessing the worker directly to bypass the WAF rule son Cloudflare
white_viel
·hace 5 meses·discuss
update: the bot is back now with a vengeance, sending request at about 1 request per second. ignoring robots.txt and the status code 403
white_viel
·hace 5 meses·discuss
serving a zip bomb and after 10 minutes, the traffic from the gpt bot disappeared..
white_viel
·hace 5 meses·discuss
will be serving a zip bomb to the bot to see if they stay away from my proxy
white_viel
·hace 10 meses·discuss
Jellyfin - an almost direct replacement of plex Emby - IMOO the best replacement if Plex is not viable