MathLogic / C#: Calculate likelyhood that login sessions are from different systems

Question

We have lots of users with a lot of userssessions. We want to determine if the software is being used on different machines. We want to calculate a likelyhood with different variables we created. So the likelyhood that it is being used on another pc:

the variables we use are things like ip, and other variables.

but lets say the user has a dynamic ip adress. the Ip adress change is very high but the other variables can stay constant. only if all variables change we want to create a bigger likelyhood that it is being used on another system.

Does anyone know a formula or implemtation for calculating this likelyhood?

thx matthy

Answer 1

This question is looking for a very specific answer given incredibly vague details, but here's my thoughts:

• It is less likely that you will have the same user log in from two IP Addresses within a short range of time without it being multiple computers. Especially if this happens frequently.
• You might try giving each variable you have some sort of weight metric that you can use as either a multiplier or constant value. Run an analysis over your logs or wherever you have the data to come up with a weighted total of all of the variables.
• Keep in mind that you can have some variables or scenarios that make the chance less likely and help offset other variables.
• Pick a threshold of when it would be likely that, given the calculated value, the users are logging in from multiple machines.

This isn't really something you can just kind of guess at. You'll likely need to do some scenario testing and collect statistics. Based on those statistics you can try the weight method and figure out a reasonable threshold and react accordingly.