models/fhe_model.py

"""
Module for training and deploying an FHE-enabled
Random Forest model using Concrete ML.
"""

import os
import pandas as pd
import joblib
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from concrete.ml.sklearn.rf import RandomForestClassifier
from concrete.ml.deployment import FHEModelDev

# Load the data (100,000 rows only)
DATA_PATH = os.path.join(os.path.abspath(os.getcwd()), "dataset", "card_transdata.csv")
df = pd.read_csv(DATA_PATH, nrows=100000)  # Limit to 100,000 rows

# Check for missing values
if df.isnull().sum().any():
    df = df.dropna()

# Handle class imbalance
fraud = df[df["fraud"] == 1]
non_fraud = df[df["fraud"] == 0].sample(n=len(fraud), random_state=42)
balanced_df = pd.concat([fraud, non_fraud])

# Separate features and target
X = balanced_df.drop(columns=["fraud"])
y = balanced_df["fraud"].astype(int)

# Split into training and validation sets
X_train, X_val, y_train, y_val = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

# Preprocessing: scale the data
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_val_scaled = scaler.transform(X_val)

# Save the scaler for later use
SCALER_PATH = os.path.join(os.path.abspath(os.getcwd()), "models", "scaler.pkl")
joblib.dump(scaler, SCALER_PATH)

# Train the Random Forest model with Concrete ML
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train_scaled, y_train)

# Compile the model for homomorphic encryption
model.compile(X_train_scaled)

# Save the model and necessary files for client and server
FHE_DIRECTORY = os.path.join(os.path.abspath(os.getcwd()), "models", "fhe_files")
dev = FHEModelDev(path_dir=FHE_DIRECTORY, model=model)
dev.save()

print("Model trained, compiled, and saved.")