There are platforms now using temporary email addresses to register bots.
The pattern of these bots is that they are all unrated and the username is a random number or letter of length 10, and they will submit several times per minute.
I suggest Mike send an email every 200 submissions to all unrated accounts' email to verify this type of bot.
Here are some examples:
0n9nm0y6ho submitted more than 60 submissions in just 20 mintues.
tzqymu69i6 50 summissions in half an hour.
zunbgg8w3g also 60 submissions in half an hour.
I want to know why they are all using Python and have the following code content:
if __name__ == "__main__":
main()
Are they training AI?
These bots seriously affect normal user code submission.
MikeMirzayanov, please ban them.
UPD： They're back. This time it's a random string followed by a random number.
But they can register many accounts and let each account submit 200 times, then use a new account.
Well, yes. It is difficult to find a fundamental solution to the bot problem. After dropping a batch of bots, new bots will always appear on CF.
Maybe that's time to verify using phone number?..
This is like
fun main() {...}(Kotlin). It will execute when you run that particular module only, but will not run if you import it to some other moduleWhen I use LLM to generate some Python code, LLM always adds this code. I have never seen anyone around me have the habit of writing code like this.
This kind of code is essentially to prevent the program from running multiple times, but it seems very redundant in algorithm competitions
It's a common Python template, not indicative of AI on its own
This isn't strange, and it's actually very normal Python code; I use this pattern in my Python template.
By using this pattern, you can make sure everything in the top level is defined before entering main(). Therefore, you can put your main() function in the top of your code (just below imports), and put auxiliary functions below main(). I feel like this is the most readable code structure and helps debugging, because you can read the code from the top and just keep going down as you read, instead of going back and forth.
Yeah, it's mostly used in Templates
To tell the truth,you can use a better chptcha system to protect CF from these robot,or just ban temporary email.Also,if do like you say,they can register new account every 200 times.
I've made a bot list and this seems to be type II, how do you know if they are using temporary email?
I have not found any place where I can register a permanent email without using my phone number. Phone numbers are limited, so it is reasonable to assume that they are not enough to register a large number of bot accounts. I have seen at least 100 bot accounts like these.
I guess a company host a email by itself and use it's email instead of temporary email.
Anyway, this should be banned.
Yeah, now there are 1/3 of these bot submissions on each page's submission record.
Companies that do this should really be banned from using CF.
Banning
Great.
But they can make their own email server and buy a domain name, then all the email addresses whose end is @<their domain name> can be used by them to register codeforces account.
so why not ban their registeration ip address
Because of VPNs. AI companies are probably smarter than you and me, anything that we can think of, they will have a way to bypass.
But aren't VPN's proxies limited
Yeah that does make sense, but it is still very hard to discern multiple accounts. The computer room of our school shares a single IP, so even if two accounts are from the same IP, it would still be unfair to ban them.
Then real users who use VPNs (like me) will be unable to submit.
Right now some people are suggesting submission bans or other some criteria when registering. However, the most extreme example is registering an account (with temporary email and VPN to prevent multiple account detection), then using it for a single day and then abandoning it. With good logistics and account naming, the testing can be spread among many accounts and become virtually undetectable. Anything short of camera monitoring will fail, so this is extremely difficult to solve.
I suggest making it that way that you can submit only like 10 submissions in every 10 minutes
Corner case :)
The problem with this is that legit users might make tons of submissions while doing a problem and far exceed this rate, but then comes a long time with no submissions at all (when eating, sleeping or thinking about a problem), while bots can simply submit 0.9 time per minute and bypass this while adding far more load to the servers than legit users. Also multiple accounts exist.
Then maybe we now need to registrate with phone number
A quick search shows that apparently temporary phone numbers exist too. Is this true?
Auto comment: topic has been updated by ink65536.