Announcing the new spam detector.
■ このスレッドは過去ログ倉庫に格納されています
Our new spam detector will detect spam in real time.
You can see the spam detector here:
http://stream.bbspink.com/spamindex.html
The page is regenerated every time the spam detector reads 10 new post replies.
You need to refresh the page manually.
The spam detector displays the top-ten spam candidates.
If there is no spam, then it displays the ten most recent posts.
"Total seen posts" is how many posts the spam detector has analyzed since the last reset.
The spam detector will periodically reset back to 0 total seen posts.
The "spam index" is a number assigned to each post. If the spam index number is 0 or 1, then that post is not spam.
The higher the spam index number, the more likely a post is to be spam.
The spam index number will slowly decrease back to zero if the spam stops.
Here is an example of spam that has been detected by the Spam Detector:
http://i.imgur.com/d5sFRLN.png
The spam detector is still being improved.
You can expect the detector algorithm to become better at detecting spam in the future.
If you want to help, then you can watch the spam detector and report any spam to the housekeeping board. I will be working on the spam detector and restarting it periodically today. The spam detector will not display posts with a 0 or 1 spam index anymore.
If there is no detected spam, then it will say, "No Spam Detected".
http://i.imgur.com/FsvaJ0w.png I just added a post locator. You can now click "Locator" and retrieve a list of URLs where that specific spam has been posted.
The list of URLs will be reset once per day. >>6
It is working well. It even finds troll posts. However it currently does
not do anything with the troll postings. It is ok for now that Jim-san happen to know there are lots of trolls
everywhere in pink.
And it must be obvious by now that more than 80% of the total spams are
posted by a one particular spammer who uses (ad-55.com/party-boy/
secret-g.net/hpyvv.com/merry2.info/mankees.com/moonshining.net/
finderz.biz and so on) I hope there will be some kind of
preventive measure to come up with.
(voiding the spammer's ●, help bringing back ROCK system which
prevent particular word from posting, regulating spammer's ISP, etc) Have you brought that up on the ccc board?
>>8 >>8
Is Rock system disabled again? >>9
No, because ccc is not functioning. Jack-san was not willing
to help even against spam.
example
Tried ccc for spam reporting
http://pele.bbspink.com/test/read.cgi/ccc/1380627185/
Refusal by Jack-san. No advice whatsoever.
http://pele.bbspink.com/test/read.cgi/erobbs/1372645155/84
Jack: "I am not going to look at it at that point"
Final
Solution: Asked Jim-san to register the word to NG-word list and the case was closed.
http://pele.bbspink.com/test/read.cgi/erobbs/1378224674/595,596,598 >>10
Old ROCK system is still working, although I think behavior of it between pele and kilauea
server is somewhat different (kilauea server accept some NG word which pele server doesn't.
Maybe looking at different ROCK NG word list? It may mean it lost NG word list fetch feature
from 2ch central ROCK server now)
2ch ROCK system now lacks new NG word registration capability and only relying
on same old NG word list. The reason for it, I heard, was the developer
could not log in to the certain 2ch server after the registration system crash
and unable to re-construct the system. >>12
We are working on something else. We will make Pink good. At this time
I can't help 2ch. >>13
Thank you for the reply. Since the spammer is posting spams
greater than consecutive posting limit and faster than consecutive
posting time limit, he/she must be using ● to break thru the
regulation. It may be good to void his ● ID to slow him down a bit, I think.
Anyway, I will be looking forward to "something else", if ISP regulation,
ROCK modification are not feasible. 200-300 spam posts 16 hours a day to specific
boards are just getting too much. Rock54は更新が止まっているだけです。
bbs.cgi内の"sub Is_Koukoku"が呼び出されている限り、古いNGWordリストを参照し続けています。
しかも効率の良くない方法で参照しているので、Is_Koukokuを呼ばないようにするのも手かもしれません。
たしか、名前、メール、本文を結合してregexpしているかと思います。
そしてNGwordはメタ文字で描かれています。 >>15
I was able to explain this to Codemonkey. Bells rang, steam engine began to move
and encoding begins. Thanks for the hint. >>15
Thank you. I am working on this now. My internet was broken all weekend. I will continue work on this now. >>15
What is the purpose of the md5 checksum in the Rock54.txt file? I have figured out what to do. I will begin implementation for it tomorrow. Now I will take a rest. Have a good evening. The new anti-spam system is now online for Pele.
If the spam detector detects spam over a certain spam-index threshold, then that spam will be added to the Rock54.txt file for `x` amount of minutes.
After the certain amount of minutes is complete, then it checks if the spam is still over the threshold or not.
If the spam is still over the threshold, then it will remain on the Rock54.txt for another `x` amount of minutes.
The spam-index threshold and amount of minutes to stay on Rock54 can be easily changed in the future.
Once spam is on the Rock54.txt, then it is business as usually for the Rock54 system.
If you encounter a bug with the system, please tell me on this thread. Have a good day! Installed on Kilauea now. >>23
It is starting to be an effective tool.
Thank you. To our unfortunate, the spammer started to change IP address between
every post since this morning, which void regular ROCK54.
http://pele.bbspink.com/test/read.cgi/erobbs/1390815733/255-256
hmmm....any possibility for SAKURA?
Note:) Regular Rock54 allows registered words to be posted twice or
three times from a given IP address in a day before blocking further
posting. SAKURA is the special registered word in Rock54 which never
allow the word to be posted even once. I have no idea that the Sakura
word is described in the Rock54 word lists though. Correction:
Wrong: I have no idea that the...
Right: I have no idea how the...
Maybe advice from ◆Rock54hC3G0C^-san as a developer of Rock54? 既に削除されていますね。
Those contents already had been deleted... >>27
元々のNGワードが判れば、Rock54リストを眺めることは出来ます。
I can be searching a Rock54-list if the original word is known.
そういえば、BBRが動いていないのだった。。（BBXは動いていると思います）
When saying so, BBR was not moving.(I thinking that BBX is moving), >>29
とりあえず削除前に必死は拾っていたみたいですが…
ttp://hissi.org/read.php/hneta/20140131/WStlYVJNZnAw.html
http://hissi.org/read.php/pub/20140131/TFgyeGZlOEow.html ふむ。
手動で更新掛けてみる？（但し反応するのはbbspinkだけだったはず） にゃあ板でテストするために登録してそうなワードはどうでもいいとしても
ソープ板で埋め立てしてる人の分を逐次登録してそうだったり
何かさわってありそうだからやめた方がいい気が…
って書き込もうとしてたら失敗した後でしたか >>27
I have already implemented SAKURA words. My apologies. Something bad happened. I am investigating the problem now. At least we all know SAKURA words for Rock54 works correctly now. Sorry about that! Everything is fixed and back to normal now. Sorry about the inconvenience. >>34
Thank you for your great effort.
Somehow the spammer got it thru still.
http://pele.bbspink.com/test/read.cgi/erobbs/1390815733/290
(Remaining 30 posts out of 120 spam postings in an hour. Other
90 posts were already deleted.) >>38
Although, this one spammer managed to get through, there have been many spammers who have not been able to get through.
I am still tweaking the settings and fixing bugs in the new spam software. As the software matures, less spammers will be able to get through. >>38-39
It is a cat and mouse game now. >>39
Thank you for the comments. I wonder why the spammer gets thru...
I noticed one behavior difference, he was posting to pub/hneta/kageki/kageki2/
pinkqa/sureh/pinkcafe boards mainly until late yesterday.
Today he posted 300 posts to pub board only. I suppose the Rock54 behavior
difference between kilauea and pele server exist? I hope you could check
for it, if you have time. ■ このスレッドは過去ログ倉庫に格納されています
