Linear regression on housing data with blind inference

[oceans404]

Show and tell project type

Novel Nada Program Example

Github Repo Link

https://github.com/oceans404/nada-linear-regression-housing/tree/main

Video Walkthough Link

No response

Project Description

I ran blind inference on Nillion to predict the house price of a new input home based on 12 features: area, bedrooms, bathrooms, stories, mainroad, guestroom, basement, hotwaterheating, airconditioning, parking, prefarea, furnishingstatus. I trained the model with this housing dataset: https://www.kaggle.com/datasets/yasserh/housing-prices-dataset/data

Result of running main.py with a new input home:

• 

What problems does your project solve? How does it preserve privacy for users?

Inference is the process of using a trained model to make predictions or decisions based on new, unseen data. It is the phase where the model is applied to input data to generate an output, such as a prediction or classification.

This is an example of blind inference, where the party running the price prediction with the new input never sees the trained model state. Instead, the trained model state is provided to the Nada program as a secret, which keeps model state private from the user running inference.

How does the project use Nillion? Describe and link to any Nada programs

Here's my linear regression nada program linear_regression_12.py, inspired by the nada-ai linear regression example. My program has

• Precision specification: na.set_log_scale(32)
• A feature count of 12, which corresponds to the 12 housing features I used to train my model feature_count = 12

# Logic from nada-ai linear_regression examples
# https://github.com/NillionNetwork/nada-ai/blob/main/examples/linear_regression/src/linear_regression.py

import nada_numpy as na

from nada_ai.linear_model import LinearRegression

def nada_main():
    # Set precision
    na.set_log_scale(32)
    # Step 1: We use Nada NumPy wrapper to create "Party0" and "Party1"
    parties = na.parties(2)

    # Step 2: Instantiate linear regression object
    # Using a Housing dataset with 12 features
    # https://www.kaggle.com/datasets/yasserh/housing-prices-dataset
    feature_count = 12
    my_model = LinearRegression(in_features=feature_count)

    # Step 3: Load model weights from Nillion network by passing model name (acts as ID)
    # In this examples Party0 provides the model and Party1 runs inference
    my_model.load_state_from_network("my_model", parties[0], na.SecretRational)

    # Step 4: Load input data to be used for inference (provided by Party1)
    my_input = na.array((feature_count,), parties[1], "my_input", na.SecretRational)

    # Step 5: Compute inference
    # Note: completely equivalent to `my_model(...)`
    result = my_model.forward(my_input)

    # Step 6: We can use result.output() to produce the output for Party1 and variable name "my_output"
    return na.output(result, parties[1], "my_output")

Is there anything else you want to share?

No response

Optional - Link your project and team members' social handles

https://twitter.com/0ceans404

Optional - Team ETH Address(es)

• 0xD3b77ed3E09B7f551B6f7b1Dd9f9e55C25D67371