With Redis supporting all of the use cases for fraud detection we could think of, we thought - why not go ahead and use the supporting features in Redis alone to solve one of the major problems that exists on the networks and web.
What it does
Project used Redis' data structures and modules to detect fraud in real time and generate insights.
How we built it
The core of the system consists of Fraud Detection Module, where different data structures and Redis modules are used. Figuring out Ad stacking used Sorted Set, Malicious IPs and device identification is through Cuckoo filter, Streaming using Redis Steams, Pub Sub, Redis gears for insight generation.
Challenges we ran into
Platform compatibility of module binaries on mac OS. Less documentation for Redis Gears and the new ones ( Now I am nitpicking).
Accomplishments that we're proud of
For a system that can handle requests of upto millions per second, having to deal with heterogenous technologies due to their inability to satisfy use cases is a huge pain. Super happy to have replaced Redis with all of them.
What we learned
The new features in Redis, implementing them is fun too.
What's next for Real time Fraud Detection
Support to improve fraud detection techniques using RedisAI.