/* Bot Spam and IP Bans */
- nicksen782
- Posts: 714
- Joined: Wed Feb 01, 2012 8:23 pm
- Location: Detroit, United States
- Contact:
Re: /* Bot Spam and IP Bans */
I wasn't sure until that last post. I'll remove that user.
- nicksen782
- Posts: 714
- Joined: Wed Feb 01, 2012 8:23 pm
- Location: Detroit, United States
- Contact:
Re: /* Bot Spam and IP Bans */
An update:
The spam bots have returned. I turned the counter-bot (I Hate Spam) on to do some deletions. I had not had a need to use it for a while since the spam has been minimal. I had also started a new version of I Hate Spam but had to stop work on it since I didn't have much in the way of test cases.
Well ... the spammers are back and test cases exist. I've rebuilt the bot and cleaned up the code and added new features too.
Features:
Happy Uzebox-ing!
The spam bots have returned. I turned the counter-bot (I Hate Spam) on to do some deletions. I had not had a need to use it for a while since the spam has been minimal. I had also started a new version of I Hate Spam but had to stop work on it since I didn't have much in the way of test cases.
Well ... the spammers are back and test cases exist. I've rebuilt the bot and cleaned up the code and added new features too.
Features:
- Never deletes posts made by a "trusted member" username.
- Can delete posts based on the user name of the poster.
- Can delete posts based on the number of "spammy words" in the topic title.
- Can now detect "Mass posts" where the post count over a span of time is too high.
- There is also an auto-add known spammer feature for certain deletions. Mass posts and topics with a higher number of spammy words are auto-added to the known spammer list.
- Can run via a cron job (meaning the server itself runs the code without a web browser.) It runs every 5 minutes.
- Has a web interface that allows easy individual post deletions, addition of trusted members and untrusted members. It also shows the last time that the cron job ran.
Happy Uzebox-ing!
Re: /* Bot Spam and IP Bans */
Thank you for keeping this in check! It's always disheartening to view the forum and get excited for some new posts only to find it's just spam.
I've been pretty busy lately with work stuff so I haven't had much time to Uzebox, but I do check this forum every day to keep up with developments.
I've been pretty busy lately with work stuff so I haven't had much time to Uzebox, but I do check this forum every day to keep up with developments.
Re: /* Bot Spam and IP Bans */
Great work man!
Re: /* Bot Spam and IP Bans */
Is it really running every 5 minutes?
Re: /* Bot Spam and IP Bans */
I'm leaving the current spam for now. These ones would be difficult to detect in software I think.
- nicksen782
- Posts: 714
- Joined: Wed Feb 01, 2012 8:23 pm
- Location: Detroit, United States
- Contact:
Re: /* Bot Spam and IP Bans */
Yes it is. It runs as a cron job set for every 5 minutes. I still have to check periodically to make sure it is working and to add new known spammers but the process is much easier now.
What do you mean? I am flagging Cyrillic text as the topic title as spam.
Re: /* Bot Spam and IP Bans */
Ah okay. The spam I was seeing must not have been from known spammers.
- nicksen782
- Posts: 714
- Joined: Wed Feb 01, 2012 8:23 pm
- Location: Detroit, United States
- Contact:
Re: /* Bot Spam and IP Bans */
Yeah, it deleted posts based on: known spammers, Cyrillic text, and the count of "spammy" words in a topic title, and additionally if a user posts too many times within a minute.
So, I check it periodically to see if it is working. I usually add a new known spammer manually each day and also do a user prune of spam accounts once identified. Feel free to PM me for more details and how to use it.
So, I check it periodically to see if it is working. I usually add a new known spammer manually each day and also do a user prune of spam accounts once identified. Feel free to PM me for more details and how to use it.
Re: /* Bot Spam and IP Bans */
What about A.I. chat bots that look at thread posts and then construct odd almost-on-topic posts that hit the uncanny-valley?
They're getting pretty darn good, it's scary. They're used in marketing a lot now. Like Google's Dialogflow.
( I've worked on Cover Girl's Olivia Virtual Greeter, and a few other commercial interactive chat-bot projects.)
The difficult/fun part in figuring them out is asking questions that don't include hint keywords that would allow advanced spam bots to construct a seemingly real answer and also not cause the bot to detect being interrogated.
Like for example asking "Where did you buy your Uzebox?" a bot can detect "<where> <buy> <PRODUCT_NAME>" and do a google shopping search for "Buy <PRODUCT_NAME>" and use the first result "adafruit.com" to give out a human-like answer.
Where I'm getting at is these posts seem too generic:
forums/search (dot) php
?author_id=13088&sr=posts
I broke the link to avoid triggering an auto-detection, might cause a Mechanical Turk mode and request a human intervention behind the scene to keep the facade. (yes, they do that too. At that point you're chatting with a real human making it very difficult to weed out fake accounts).
That could still be a real person but it just seems oddly bland for this forum.
They're getting pretty darn good, it's scary. They're used in marketing a lot now. Like Google's Dialogflow.
( I've worked on Cover Girl's Olivia Virtual Greeter, and a few other commercial interactive chat-bot projects.)
The difficult/fun part in figuring them out is asking questions that don't include hint keywords that would allow advanced spam bots to construct a seemingly real answer and also not cause the bot to detect being interrogated.
Like for example asking "Where did you buy your Uzebox?" a bot can detect "<where> <buy> <PRODUCT_NAME>" and do a google shopping search for "Buy <PRODUCT_NAME>" and use the first result "adafruit.com" to give out a human-like answer.
Where I'm getting at is these posts seem too generic:
forums/search (dot) php
?author_id=13088&sr=posts
I broke the link to avoid triggering an auto-detection, might cause a Mechanical Turk mode and request a human intervention behind the scene to keep the facade. (yes, they do that too. At that point you're chatting with a real human making it very difficult to weed out fake accounts).
That could still be a real person but it just seems oddly bland for this forum.