StopForumSpam confidence levels above 100%

[Ofnuts]

Looking at the SFS logs of my forum, it appears that since May 17th the confidence level (as logged) are about the double of what they used to be (when from 90-100% to 180-200%).

At the same time the MyBB settings only allow a max confidence level of 100%, so it could be rejecting legit users that would have a "real" SFS confidence level slightly above 50%.

Am I the only one to experience this? Is there a fix?

Posted a similar question on the SFS forum but didn't get a pertinent answer....

[Euan T]

Good catch. Total confidence can be up to 300% (it's username confidence + email confidence + IP address confidence). That total confidence should then be divided by the number of checks to get a value 0-100 rather than 0-300. Ive opened an issue on GitHub here: https://github.com/mybb/mybb/issues/2910

[Ofnuts]

Good catch. Total confidence can be up to 300% (it's username confidence + email confidence + IP address confidence). That total confidence should then be divided by the number of checks to get a value 0-100 rather than 0-300. Ive opened an issue on GitHub here: https://github.com/mybb/mybb/issues/2910

But then why did it change? I don't remember changing the MyBB code in that time frame.

[Euan T]

Good catch. Total confidence can be up to 300% (it's username confidence + email confidence + IP address confidence). That total confidence should then be divided by the number of checks to get a value 0-100 rather than 0-300. Ive opened an issue on GitHub here: https://github.com/mybb/mybb/issues/2910

But then why did it change? I don't remember changing the MyBB code in that time frame.

There hasn't been any changes in MyBB relating to StopForumSpam for quite a long time now. It's possible that the actual 3rd party API has changed slightly, but I haven't looked at the API documentation since I first wrote the core functionality to handle StopForumSpam so I cannot say for sure.

[laie_techie]

I'm not sure if it makes more sense to average the 3 values or to take the max.

username: 10%
IP address: 75%
Email: 30%

Which should be the overall confidence:
* 38% (average)
* 75% (max)
* 10% (min)

[Euan T]

Yes, that's another thing that needs to be worked out.

Emails should obviously have more weight than IP (which can be dynamic) and username (which can happen by chance). Quite qhat the best formula is, I'm not so sure.

[Ofnuts]

Yes, that's another thing that needs to be worked out.

Emails should obviously have more weight than IP (which can be dynamic) and username (which can happen by chance). Quite qhat the best formula is, I'm not so sure.

Since we can already select the criteria, a logical extension is to replace this "binary" choice with a 0-100% weight.

[Euan T]

That would work, yes. I'll look at doing that for 1.8.14 this week.

[metulburr]

Why not just abolish the overall, and show/handle each category. The admin can choose which on which level of each to consider spam?

[Euan T]

That is the plan.