feat(prediction): add prediction model for l2 and improve report

src/agents/l2_supervisor.py

@@ -0,0 +1,56 @@
+import joblib
+from fastapi import HTTPException
+
+class Supervisor:
+    def __init__(self, model_path: str = "src/models/L2_XGBOOST_Supervisor_V0.joblib"):
+        try:
+            saved = joblib.load(model_path)
+            self.model = saved["model"]
+            self.scaler = saved.get("scaler", None)  # certains modèles peuvent ne pas avoir de scaler
+            self.features = [
+                "Header_Length",
+                "Time_To_Live",
+                "Rate",
+                "Tot sum",
+                "Tot size",
+                "Min",
+                "Max",
+                "AVG",
+                "Std",
+                "Variance",
+                "IAT",
+                "Number",
+                "syn_ratio",
+                "ack_ratio",
+                "fin_ratio",
+                "rst_ratio",
+                "mean_pkt_size",
+                "pkt_size_range",
+                "pkt_size_ratio",
+                "mean_iat",
+                "pkt_rate",
+                "throughput",
+                "bytes_per_sec",
+                "coef_var",
+                "tcp_udp_ratio",
+                "flag_entropy",
+            ]
+        except FileNotFoundError:
+            raise HTTPException(status_code=500, detail=f"Modèle '{model_path}' introuvable")
+        except Exception as e:
+            raise HTTPException(status_code=500, detail=f"Erreur lors du chargement du modèle: {e}")
+
+    def predict(self, data):
+        try:
+            # Préparer les features
+            X = data[self.features]
+
+            # Appliquer le scaler si existant
+            if self.scaler is not None:
+                X = self.scaler.transform(X)
+
+            # Prédictions
+            preds = self.model.predict(X)
+            return preds.tolist()
+        except Exception as e:
+            raise HTTPException(status_code=500, detail=f"Erreur lors de la prédiction: {e}")

src/controllers/prediction_controller.py

@@ -1,7 +1,10 @@
 from fastapi import HTTPException, UploadFile
 from src.services.preprocess_features import preprocess_features
 from src.services.file_verification import file_verification
+from src.services.report import summarize_predictions
+
 from src.agents.l1_screener import Screener
+from src.agents.l2_supervisor import Supervisor
 
 def global_prediction_on_csv(file: UploadFile):
     try:
@@ -15,18 +18,16 @@ def global_prediction_on_csv(file: UploadFile):
         print("start loading model and predicting")
 
         screener = Screener()
-        predictions = screener.predict(data)
+        l1_summary = summarize_predictions(screener.predict, data)
+        
+        supervisor = Supervisor()
+        l2_summary = summarize_predictions(supervisor.predict, data)
+        
 
-        total_samples = len(predictions)
-        total_attacks = sum(predictions)           # 1 = attaque
-        total_normal = total_samples - total_attacks  # 0 = requête normale
         return {
-            "message": "Prédictions effectuées avec succès",
-            "total_samples": len(predictions),
-            "total_attacks": total_attacks,
-            "total_normal_requests": total_normal,
-            "predictions": predictions
-        }
+                "l1": l1_summary,
+                "l2": l2_summary
+                }
 
     except HTTPException:
         raise

src/models/L1_Logistic_v0.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:79d5f16e998f61fdf1faff94e137865437cf932fc6a1fa5c6b01fa3cf4302c47
-size 2617
+oid sha256:2a84679401d8c2c9fb18f98cac36aba8fe6638f5baee9c10584199adfe504c0b
+size 2605

src/services/report.py

@@ -0,0 +1,39 @@
+import time
+from typing import List, Dict, Callable
+
+def summarize_predictions(pred_func: Callable, data) -> Dict:
+    """
+    Exécute une fonction de prédiction sur des données et retourne
+    les statistiques générales des prédictions.
+
+    Args:
+        pred_func: fonction de prédiction, doit accepter `data` et retourner une liste ou array de prédictions binaires (0/1)
+        data: données sur lesquelles effectuer la prédiction
+
+    Returns:
+        dict contenant : total_samples, total_attacks, total_normal,
+        processing_time, avg_pred_time, attack_ratio, normal_ratio
+    """
+    start_time = time.time()
+    predictions = pred_func(data)
+    end_time = time.time()
+
+    total_samples = len(predictions)
+    total_attacks = sum(predictions)          # 1 = attaque
+    total_normal = total_samples - total_attacks
+
+    processing_time = end_time - start_time
+    avg_pred_time = processing_time / total_samples if total_samples > 0 else 0
+    attack_ratio = total_attacks / total_samples if total_samples > 0 else 0
+    normal_ratio = total_normal / total_samples if total_samples > 0 else 0
+
+    return {
+        "total_samples": total_samples,
+        "total_attacks": total_attacks,
+        "total_normal": total_normal,
+        "processing_time_sec": processing_time,
+        "avg_pred_time_sec": avg_pred_time,
+        "attack_ratio": attack_ratio,
+        "normal_ratio": normal_ratio,
+        "predictions": predictions
+    }