Final project for the Building AI course
An app that detects irregular behavior and warns users to be alert for scams
The problem is that many people are or will be the victim of online fraud, a billion dollar industry, because they either:
- lack awareness
- are to distracted and busy with other things to be alert
- the complexity and quality of online fraud becomes so good its hard recognize
This app will mitigate the alertness factor, its does not scan the hardware for a virus but rather 'scans' the user behavior for irregularities that might be classified as an online fraud attempt on the user. Making the user aware by pointing out the flagged behavior gives him/her the time to re-evaluate.
As an application running in the background, initial target users will be office workers, elderly and young people.
Tricky part which requires more thought:
- How to train an AI to detect outliers (e.g. a never seen before pattern)
- data will be come from users as their behavior is partly generic and personal, will need to be anonimized or locally trained
It does not solve online fraud
I need technical skilled people in data science/engineering, appropiate datasets and test groups to make and deploy a minimal viable model and iterate from there