Inspiration
Amazon guard duty
What it does
Analyze the user login logs, and monitors and predicts whether there is any suspicious activity
How I built it
I have used logistic regression with hand picked features to detect anomalies. It is built using sklearn and general python ecosystem
Challenges I ran into
location api is bottleneck and so we are caching the results. Scalability can become an issue when there is huge influx of logs
Accomplishments that I'm proud of
We have deployed it in our internal product and got very positive feedback
What I learned
Security practices are still not followed even in very reputed firms.
What's next for log anomaly detection
Exploring more features and scaling it using spark or other cluster based solutions
Log in or sign up for Devpost to join the conversation.