| Author |
 Topic  |
|
AnonJr
Moderator
United States
5768 Posts |
|
|
Panhandler
Average Member
USA
783 Posts |
 Posted - 11 March 2008 : 21:21:11
|
quote:
Originally posted by Podge
There are ways around everything.
Podge, what if the question is a random generated number and the added number is also random generated (but kept small enough that it would be simple).
???
Machine vs machine?
|
"5-in-1 Snitz Common Expansion Pack" - five popular mods packaged for easy install
". . .on a mote of dust, suspended in a sunbeam. . ."
HarborClassifieds
Support Snitz Forums
|
 |
|
|
phy1729
Average Member
USA
589 Posts |
Posted - 11 March 2008 : 21:32:34
|
A program could be made that assigns probabilities to all fields and the spammyness if words in it like http://www.paulgraham.com/spam.html . Or some person is going to create a company that will create all these questions and create a new industry. Intellectual property in the previous sentence is mine.  |
Edited by - phy1729 on 11 March 2008 21:32:51 |
|
|
|
Podge
Support Moderator
Ireland
3775 Posts |
|
|
HuwR
Forum Admin
United Kingdom
20584 Posts |
Posted - 12 March 2008 : 08:20:31
|
definitely needs some work, since probablity scores should lie between 0 and 1, many of yours are > 1  |
|
|
|
Podge
Support Moderator
Ireland
3775 Posts |
|
|
HuwR
Forum Admin
United Kingdom
20584 Posts |
Posted - 12 March 2008 : 10:34:43
|
the scores probably just need rounding to 2 decimal places which would probably eliminate scores over 1 and under 0.
probably that degree of database traffic isn't best suited to Access |
|
|
|
pdrg
Support Moderator
United Kingdom
2897 Posts |
Posted - 12 March 2008 : 11:16:53
|
My proposal (with all the limitations of CAPTCHA's, fair enough) is to use serverside imagemaps as a solution -
'click on the yellow circle'
'click on the smiling face'
'click on the symbol that looks like this: <some shape>'
The images can be created dynamically, or created in bulk in a library of gifs, as long as the range of 'good' x and y coordinates are known, you have a CAPTCHA-style challenge with the added complication of being click-based (so not as easy as passing the CAPTCHA through to another site to be solved).
Is it accessibility-friendly? Not at all, so it requires an accessibility manual approval option too. But CAPTCHA's aren't either. It's an option anyway. |
|
|
|
HuwR
Forum Admin
United Kingdom
20584 Posts |
Posted - 12 March 2008 : 11:21:54
|
| images are a nono for accesibility reasons, blind readers do not display images |
|
|
|
Podge
Support Moderator
Ireland
3775 Posts |
Posted - 12 March 2008 : 11:40:43
|
quote:
the scores probably just need rounding to 2 decimal places
Its because there are more words on the forum than there are in each corpus. When I was trying to generate the good (ham) corpus it was taking too long so I stopped it. The scores are skewed towards spam considerably as a result.
I think I'll reopen the old topic rather than hijack this one.
Edit: I'll start a new topic - the old one is in a private forum. |
Podge.
The Hunger Site - Click to donate free food | My Blog | Snitz 3.4.05 AutoInstall (Beta!)
My Mods: CAPTCHA Mod | GateKeeper Mod
Tutorial: Enable subscriptions on your board
Warning: The post above or below may contain nuts. |
Edited by - Podge on 12 March 2008 11:54:01 |
|
|
|
AnonJr
Moderator
United States
5768 Posts |
Posted - 12 March 2008 : 13:21:01
|
You can keep using this one if you want... its not really off topic. The main thing I've been looking at is ways to stop spam without unduly harassing the visitors.
Since I wrote the initial article, and applied most of the things I wrote about (I don't go for IP banning or CAPTCHAs), I'd halted the bogus registrations from ending up in my pending members list ... for a time. Its now back to being as bad as it was before (same IPs as before, but again I don't really go for banning IP addys).
I'm looking for newer, better ways to thwart these ... people. I'd read a number of articles on Bayesian filtering, but some of these sites run off Access (with no access to SQL Server either) - and as mentioned earlier, they just can't handle the db load.
I'm tracking down a few other options. Unfortunately I just haven't had the time to try some of them out. I'm looking at updating my original spam fighting article sometime soon with the things I've read lately. This topic was started as a means of generating some useful input and leads - which it has. |
|
|
|
HuwR
Forum Admin
United Kingdom
20584 Posts |
Posted - 12 March 2008 : 13:56:49
|
| well, for some reason we don't seem to get that many bogus registrations here |
|
|
|
AnonJr
Moderator
United States
5768 Posts |
Posted - 12 March 2008 : 14:43:13
|
I'm not seeing the same volume over at www.jesusjoshua2415.com nor at www.eastoverfd.com - but for some reason the church website is just getting hammered. There are over 50 pending members - all would-be spammers. 
I think I've had a total of 5 (combined) at the other two sites. =/ |
Edited by - AnonJr on 12 March 2008 14:43:34 |
|
|
|
HuwR
Forum Admin
United Kingdom
20584 Posts |
Posted - 12 March 2008 : 16:29:20
|
| I am not entirely convinced that your method is as effective as you think, since most of the bots use an externally constructed form to do the post which would just bypass the fact you had swapped the buttons |
|
|
|
phy1729
Average Member
USA
589 Posts |
Posted - 12 March 2008 : 16:48:07
|
My log would indicate that the bot does check policy.asp
2008-03-06 22:47:53 W3SVC5251 DEDHSTWE_TEMP *ip.add.re.ss* GET /forum/policy.asp - 80 - 82.135.148.179 HTTP/1.1 - - - my.url.example.com 302 0 0 445 60 218
Complete log: http://phy1729.googlepages.com/nestolam.log |
|
|
|
Topic |
|
Topic Locked
. The problem lies in generating the corpi (coorpuses?). Basically, if a post has 50 words then its a minimum of 100 sql statements to insert the words into the spam or ham corpi. I need to find a faster way to update each corpus for each post. I know how to do it with MySql or MSSQL (sproc) but its impossible with Access.