Architecture of our solution
We got motivated to do this project as it lets us be innovative and provide a solution to a large scale problem.
What it does
Our solution uses blockchain to create a tamper-proof record of IoT meter readings. We also use different machine learning -namely K-means clustering, anomaly detection with autoencoders and LSTM.
How we built it
We built a node.js server on which our front-end was deployed. Queries from our front-end are passed to the blockchain via rest-api server. Anomaly detection model was run on flask server which is used to detect tampered IoT meter readings.
Challenges we ran into
Implementing an incentive model (WRCs) that incorporates different types of organizations. Coming up with a machine learning model that fits seemingly countless possibilities of fraud.
Accomplishments that we're proud of
We deployed our model on Hyperledger successfully. We were able to notice some patterns of tampering on the IoT meters to train our model for.
What we learned
We got a peek into implementing a real-life solution that affects a really large audience.
What's next for us
Once blockchain scales, our solution can be considered for deploying on a larger scale. Our machine learning algorithms can be trained for more cases of tampering.