pryvx 2.4.1__tar.gz → 2.4.3__tar.gz

{pryvx-2.4.1 → pryvx-2.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pryvx
-Version: 2.4.1
+Version: 2.4.3
 Summary: A comprehensive package for privacy-enhancing technologies
 Home-page: UNKNOWN
 Author: PryvX (Jayesh Kenaudekar)

{pryvx-2.4.1 → pryvx-2.4.3}/pryvx/__init__.py

@@ -1,4 +1,5 @@
 from .psi import OPRF
 from .phe import PHE
 from .smpc import SMPC
-from .gdp import GDP
+from .gdp import GDP
+from .ldp import LDP

pryvx-2.4.3/pryvx/fl_client.py

@@ -0,0 +1,36 @@
+import grpc
+from pryvx import pryvx_pb2
+from pryvx import pryvx_pb2_grpc
+from sklearn.linear_model import LogisticRegression
+from sklearn.ensemble import RandomForestClassifier
+import pickle
+
+
+def train_logistic_classifier(features, labels):
+    model = LogisticRegression()
+    model.fit(features, labels)
+    serialized_model = pickle.dumps(model)
+    return serialized_model, model
+
+
+def train_random_forest_classifier(features, labels):
+    model = RandomForestClassifier(n_estimators=100, random_state=42)
+    model.fit(features, labels)
+    serialized_model = pickle.dumps(model)
+    return serialized_model, model
+
+
+def send_params(serialized_model, connection_url, client_id):
+
+    with grpc.insecure_channel(connection_url) as channel:
+        stub = pryvx_pb2_grpc.ModelServiceStub(channel)
+
+        # Attach metadata (client ID) in the request
+        metadata = (("client_id", client_id),)
+
+        model_params = pryvx_pb2.ModelParams(params=serialized_model)
+
+        response = stub.SendModelParams(model_params, metadata=metadata)
+
+        return f"✅ Model Params sent to server from {client_id}"
+

pryvx-2.4.3/pryvx/fl_ensemble.py

@@ -0,0 +1,20 @@
+import numpy as np
+
+def federated_ensemble(estimators, X_test):
+    """
+    Perform soft voting ensemble on a list of trained estimators.
+    
+    :param estimators: List of trained models that support `predict_proba`
+    :param X_test: Test dataset (features only)
+    :return: Final predictions after soft voting
+    """
+    if not estimators:
+        raise ValueError("No models provided for ensemble.")
+
+    # Sum all probability predictions
+    avg_probs = sum(model.predict_proba(X_test) for model in estimators) / len(estimators)
+
+    # Convert probabilities to final predictions (highest probability wins)
+    y_pred_ensemble = np.argmax(avg_probs, axis=1)
+    
+    return y_pred_ensemble

pryvx-2.4.3/pryvx/fl_server.py

@@ -0,0 +1,51 @@
+import grpc
+from pryvx import pryvx_pb2
+from pryvx import pryvx_pb2_grpc
+import pickle
+from concurrent import futures
+import os
+import datetime
+
+# Directory to store received models
+MODEL_SAVE_PATH = "received_models/"
+os.makedirs(MODEL_SAVE_PATH, exist_ok=True)
+
+class ModelServicer(pryvx_pb2_grpc.ModelServiceServicer):
+    def __init__(self):
+        self.client_models = {}  # Store models from clients
+
+    def SendModelParams(self, request, context):
+        try:
+            # Extract metadata (client ID)
+            client_id = dict(context.invocation_metadata()).get("client_id", "unknown_client")
+            
+            # Deserialize model received from client
+            model = pickle.loads(request.params)
+
+            # Assign a unique client ID
+            client_id = f"client_{len(self.client_models) + 1}"
+            self.client_models[client_id] = model
+            
+            # Save model to disk
+            model_filename = os.path.join(MODEL_SAVE_PATH, f"{client_id}_model.pkl")
+            with open(model_filename, "wb") as f:
+                
+                pickle.dump(model, f)
+
+            print(f"✅ Received and saved model from {client_id}")
+
+            return pryvx_pb2.ModelResponse(message=f"Model from {client_id} received and saved.")
+        except Exception as e:
+            print(f"❌ Error processing model: {e}")
+            return pryvx_pb2.ModelResponse(message="Failed to process model")
+
+
+def start_server():
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
+    pryvx_pb2_grpc.add_ModelServiceServicer_to_server(ModelServicer(), server)
+    server.add_insecure_port('[::]:50051')
+    server.start()
+    print("Server started. Listening on localhost:50051")
+
+    server.wait_for_termination()
+

pryvx-2.4.3/pryvx/gdp.py

@@ -0,0 +1,27 @@
+import numpy as np
+
+# Laplace Mechanism for continuous data.
+# Gaussian Mechanism for continuous data (commonly used in GDP).
+    
+def laplace_mechanism(query_result, sensitivity, epsilon):
+    """Applies the Laplace mechanism for GDP."""
+    scale = sensitivity / epsilon
+    noise = np.random.laplace(0, scale, 1)
+    return query_result + noise[0]
+
+def gaussian_mechanism(query_result, epsilon=0.5, delta=1e-5, sensitivity=1.0):
+    """Adds Gaussian noise to a global query result."""
+    sigma = np.sqrt(2 * np.log(1.25 / delta)) * sensitivity / epsilon
+    noise = np.random.normal(0, sigma)
+    return query_result + noise
+
+class GDP:
+    @staticmethod
+    def add_noise(query_result, sensitivity, epsilon):
+        return laplace_mechanism(query_result, sensitivity, epsilon)
+    
+    @staticmethod
+    def add_gaussian_noise(query_result):
+        return gaussian_mechanism(query_result)
+
+

pryvx-2.4.3/pryvx/ldp.py

@@ -0,0 +1,26 @@
+import numpy as np
+import random
+
+# Laplace Mechanism (common for numeric data)
+# Randomized Response (common for categorical/binary data)
+
+def laplace_mechanism(value, epsilon, sensitivity=1.0):
+    """Adds Laplace noise to a numeric value for LDP."""
+    scale = sensitivity / epsilon
+    noise = np.random.laplace(0, scale)
+    return value + noise
+
+def randomized_response(value, epsilon):
+    """Implements randomized response for binary data."""
+    p = np.exp(epsilon) / (1 + np.exp(epsilon))
+    return value if random.random() < p else 1 - value
+
+class LDP:
+    @staticmethod
+    def add_numerical_noise(value, epsilon):
+        return laplace_mechanism(value, epsilon)
+    
+    def add_categorical_noise(value, epsilon):
+        return randomized_response(value, epsilon)
+
+

{pryvx-2.4.1 → pryvx-2.4.3}/pryvx.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pryvx
-Version: 2.4.1
+Version: 2.4.3
 Summary: A comprehensive package for privacy-enhancing technologies
 Home-page: UNKNOWN
 Author: PryvX (Jayesh Kenaudekar)

{pryvx-2.4.1 → pryvx-2.4.3}/pryvx.egg-info/SOURCES.txt

@@ -3,8 +3,10 @@ README.md
 setup.py
 pryvx/__init__.py
 pryvx/fl_client.py
+pryvx/fl_ensemble.py
 pryvx/fl_server.py
 pryvx/gdp.py
+pryvx/ldp.py
 pryvx/phe.py
 pryvx/pryvx_pb2.py
 pryvx/pryvx_pb2_grpc.py

{pryvx-2.4.1 → pryvx-2.4.3}/setup.py

@@ -1,6 +1,6 @@
 from setuptools import setup, find_packages
 
-VERSION = '2.4.1'
+VERSION = '2.4.3'
 
 
 # Setting up

pryvx-2.4.1/pryvx/fl_client.py

@@ -1,27 +0,0 @@
-import grpc
-from pryvx import pryvx_pb2
-from pryvx import pryvx_pb2_grpc
-from sklearn.linear_model import LogisticRegression
-import pickle
-
-
-def train(features, labels):
-    model = LogisticRegression()
-    model.fit(features, labels)
-
-    serialized_model = pickle.dumps(model)
-
-    return serialized_model
-
-
-def send_params(serialized_model, connection_url):
-
-    with grpc.insecure_channel(connection_url) as channel:
-        stub = pryvx_pb2_grpc.ModelServiceStub(channel)
-
-        model_params = pryvx_pb2.ModelParams(params=serialized_model)
-
-        response = stub.SendModelParams(model_params)
-
-        return "Model Params sent to server"
-

pryvx-2.4.1/pryvx/fl_server.py

@@ -1,31 +0,0 @@
-import grpc
-from pryvx import pryvx_pb2
-from pryvx import pryvx_pb2_grpc
-import pickle
-from concurrent import futures
-
-# Server
-class ModelServicer(pryvx_pb2_grpc.ModelServiceServicer):
-    def __init__(self):
-        self.client_params = {}
-
-    def SendModelParams(self, request, context):
-        # Deserialize the model
-        loaded_model = pickle.loads(request.params)
-
-        # save model to gcp storage bucket
-
-        print("Received model params from client")
-
-        return pryvx_pb2.ModelParams(params=request.params)
-
-
-def start_server():
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
-    pryvx_pb2_grpc.add_ModelServiceServicer_to_server(ModelServicer(), server)
-    server.add_insecure_port('[::]:50051')
-    server.start()
-    print("Server started. Listening on localhost:50051")
-
-    server.wait_for_termination()
-

pryvx-2.4.1/pryvx/gdp.py

@@ -1,15 +0,0 @@
-import numpy as np
-
-def laplace_mechanism(true_value, sensitivity, epsilon):
-    """Applies the Laplace mechanism for GDP."""
-    scale = sensitivity / epsilon
-    noise = np.random.laplace(0, scale, 1)
-    return true_value + noise[0]
-
-
-class GDP:
-    @staticmethod
-    def add_noise(query_result, sensitivity, epsilon):
-        return laplace_mechanism(query_result, sensitivity, epsilon)
-
-