Morgan Tocker ([info]mtocker) wrote,
@ 2007-12-28 01:42:00
Previous Entry  Add to memories!  Tell a Friend!  Next Entry
Entry tags:mysql, php

There's nothing like a machine doing the work for you.
So I was reading Nate's PHP tip, and it got me thinking; he's basically describing bayesian session validity checks - what an awesome idea![1]. Here are my list of possible 'items' to form the test:

  • Login IP address
  • Login location (gathered courtesy of GeoIP)
  • Operating System
  • Web Browser
  • Time of Day
  • Origin of requests (both HTTP_REFERER and REQUEST_URI)
I was thinking of how damn accurately you could identify some specific cases of fraud though:
  • "I only login from work" (time of day is always 9-5, location is always the same)

Each bit of information becomes more relevant the more frequently it occurs exclusively (or exclusively as part of a set). The only catch is that I would also need to identify the patterns that form sets:
  • "At work I use Windows+IE at home I use Mac+Safari"
(i.e. Login location is one of two places. If Windows was used from home, it's not consistent with the pattern).

If I can identify the sets, then it might become frighteningly accurate. I can then classify a login attempt into one of three classes; Spam, Unsure and Not Spam. For the not spam, if it's an open session it can continue unhindered. If I am unsure, I may ask them to log in again (which is only a minor inconvenience). If it's spam, I want them to re authenticate and solve a CAPTCHA test. I want phishing to become more expensive.

[1] Although I know Google is already doing this with my Gmail.

Create an Account
Forgot your login or password?
Login w/ OpenID
English • Español • Deutsch • Русский…