Inpuri Anti-Bot

Malicious bots and automation tools pose a threat to web applications everywhere. Inpuri is a captcha-less solution to block unwanted traffic. Using only client sensor data we use ML to block bots.

Comment 1

Inspiration/What it does

Are you tired of old captchas trying to figure out whether you're a robot? That's why we created a fully captcha-less solution to stop bots and bad actors. We can predict whether a user is legitimate by using sensor data gathered by the client (such as mouse movements, browser size, user agent, etc) paired with a machine learning algorithm that has been trained on countless sensor payloads. In doing this, Inpuri helps maintain a secure environment for customers and businesses while keeping the user experience as frictionless as possible!

How we built it

We used various technologies such as Go, MongoDB, TS, Python, scikit-learn, and more to build an end-to-end solution. Go, TS, and Mongo were used in the data collection process to gather training data from both real and bot users. We utilized Python and more specifically sklearn to implement the machine learning algorithm to analyze sensor data payloads.

Challenges we ran into/Accomplishments that we're proud of

Data collection was tough in such a short time period. Because of this, we were forced to quickly set up a collection site, distribute the link to real users, and collect data legitimate data that way. To get bot data, we set up various scripts using different browser automation libraries (especially those designed to avert anti-bot systems) and collected sensor payloads through the same site. We ended up with just under 1,000 payloads to train our model with, which worked out OK, however, more training is necessary.

What we learned

We learned a lot about technologies we were unfamiliar with, especially sklearn and machine learning.

What's next for Inpuri Anti-Bot

More training/testing for ML algorithm to get confidence scores and accuracy up.

Share this project:

Updates