Some idiot script kiddie has written a script that spams /a/ with the same image rotated by 90 degrees. at the moment it is fucking the mods over because the content is changing therefore they cant ban by ip.
What im asking is for you guys to come up with a way to combat this spam. Cheers
>>44
fast enough for what? posting gigabytes of CP?
anything less than that and you can have all three:
1. go to proxy.org
2. try random proxies
3. ???
4. PROFIT!!!
Name:
Anonymous2009-03-03 2:44
>>46
>posting anonymously
>interactive web proxies
DOHOHOHOHO
Name:
Anonymous2009-03-03 3:02
>>45,46 isn't always easy
Getting one hit out of 30 is not my idea of easy.
>>48
you only need one, and trying 30 only takes about a minute.
Name:
Anonymous2009-03-03 3:25
how does spam bot know what get new proxy is?
Name:
The Solution2009-03-03 7:40
No fucking CAPTCHA!
The solution that 4chan is searching for is Commuinty Moderation. The type that has been successful on digg and youtube. (only a little more srs)
Put a [ban user for post] button on every post. then attach a voting system to the button.
Once a set number of unique I.P.'s vote to ban the user he/she will get banned for a month, and after another set number, he/she gets banned permanently.
make the unique I.P. # high, like 20 unique I.P's, and if the User tries to use a proxy whilst banning another user he/she will get banned instead.
Spam solved
4chan saved
cancer over
Name:
FrozenVoid2009-03-03 7:47
>>51 This can be abused and misused.
It would just stifle free speech and solve nothing as Dynamic IPs exist.
________________________________
Christmas is cancelled.
>>51 The type that has been successful on digg and youtube
Neither Digg nor Youtube implement meaningful community moderation schemes.
Once a set number of unique I.P.'s vote to ban the user he/she will get banned for a month, and after another set number, he/she gets banned permanently.
Because that won't be abused, at all. Spoiler: IPs aren't a reliable way to individualize end-users.
There's nothing wrong with the current report system. It does almost the exact same thing, albeit without the "LOL LET'S GANG UP AND BAN EVERYONE ELSE" feature. Oversight is necessary.
Name:
Anonymous2009-03-03 9:09
Store the hashes of posted images in a table, along with an integer that records how many times the image has been posted. In the post script, if the image has been posted too many times already, abort.
lol futaba
Name:
Anonymous2009-03-03 9:28
>>54
The hash would have to be based on an approximation of the image itself rather than the file data, so matches can be made to images that are very similar but not identical.
Name:
Anonymous2009-03-03 9:55
>>54
Sorry, but I don't understand the concept of ``posted too many times.''
In these annoying spamful times is that most everyone likes to offer their opinion on how to solve the issue of abuse, on a website that offers content to the public on somebody else's dime and with practically no formal requirements to join -- that right there is a system in which it is just begging to be abused.
Face it, any restrictions you put upon the site will only affect legitimate users, and the spammer will get his jollies off to the fact that changes have been made to the site that make it overall more difficult for to use. These changes will actually feed his ego and he will proceed to spam the site even more in order to make the place that much more shittier for everyone else. Even Futaba Channel (2chan.net) that 4chan is based off of suffer from similar issues. I'm sure if there were other imageboard sites that became as popular as 4chan they too would suffer from similar problems.
Name:
Anonymous2009-03-03 16:25
>>57 >>1 the same image rotated by 90 degrees same image rotated by 90 degrees same image rotated by 90 degrees same image rotated same image rotated same image same image
>>61
but its a different file, with a different hash, and a different whatever, and has been re-encoded
Name:
Anonymous2009-03-03 17:51
If Lisp was a car, it wouldn't start if you forget to close any of the doors.
Name:
Anonymous2009-03-03 17:58
If Lisp was a cdr, it wouldn't start if you forget to open any of the doors.
Name:
Anonymous2009-03-03 18:00
I don't think image recognition is an option either because of cpu-usage (have to compare the same image against many previous images). Also once you get the system in place, the spammer can just add a new twist to his software which will bypass the spam-protection. There are hundreds of filters at his disposal for making his every image a bit different.
I've made a few webabusers my own (nothing related to spam), and the most interesting part about is trying to bypass someone elses bot-detection. Honestly, the only way you can really stop him is to use a good CAPTCHA. And those really aren't an option.
Name:
Anonymous2009-03-03 19:59
>>66
Hey everyone I know nothing about hashing and how it works! >>51
Hey everyone I know nothing about human nature and how it works! >>54
Hey everyone I know nothing about the 4chan duplicate image detection and how it works!
To any EXPERT PROGRAMMER the solution is quite obvious. A customized hashing algorithm that is not commutative and simply takes generalizations about the image as the hash input. i.e.: average hue of each group of 10x10 pixels (perhaps not the best example, I don't really know what hue is or how likely images are to share have the same frequency for hue values of said blocks but you get the point).
Name:
Anonymous2009-03-03 20:36
>>67 is right; in fact I think /prog/ had a previous thread about just such an image hashing algorithm.
Using hue, as suggested in >>67, doesn't really work -- you get massive clusterings of unrelated images when the image is primary white/black. A reasonable similiar-image detection system is IQDB[2] which uses diffusion tensor analysis (or something that sounds like that). References:
[1] http://dis.4chan.org/read/prog/1209700628
[2] http://iqdb.hanyuu.net/code/
>>70
I looked it up. It's diffusion HAXmyANUS analysis.
Name:
Anonymous2009-03-04 0:05
>>69
"You simply draw a rough sketch of what you want to find and imgSeek displays for you a thumbnail view of the best matches."
Does this actually work? I would be extremely interested in viewing the described algorithm; but my guess is that it is not very good- so I'm not going to waste my time just yet.
Name:
Anonymous2009-03-04 0:19
>>69
Could you not make the hash(although this is leading away from hashing quite quickly) also account for the sparsity/density of the overall hues, such that an image with more densely packed hues will need to be a closer relation to other images to be considered a match?
Name:
Anonymous2009-03-04 2:42
>>72
I presume the "algorithm" is just feature detection.
Name:
Anonymous2009-03-04 4:05
what lang would you like a solution in
Name:
Anonymous2009-03-04 7:30
HELLO EVERYONE
I read about a CAPTCHA-STYLE system, that involves clicking 3 images out of 9 that apply to a word. So it was like - "click the cats", and you'd have to select the cats and not the dogs. But a bot would find this difficult. Clicking 3 cats is easier than typing words I find.
Name:
Anonymous2009-03-04 10:41
That is the best captcha system I've ever heard of.