.gitignore

@@ -7,4 +7,3 @@ workspace.code-workspace
 .VSCodeCounter
 wrks
 tmp
-__pycache__

about.md

@@ -1,57 +0,0 @@
-# PerfSPEC Learning Phase
-
-Based in [PrefSPEC: Performance Profiling-based Proactive Security Policy Enforcement for Containers](https://ieeexplore.ieee.org/document/10577533) document presented in [1], thir repository contains source files used to generate and process data.
-
-[PrefSPEC document](PerfSPEC.pdf)
-
-[Presentación in Spanish](presentacion.pdf)
-
-<div style="margin: auto">
- <a target="_blank" href="perfspec-learning/src/branch/main/presentacion.pdf"><img src="imgs/prefSPEC-learning.png" width="800"></a>
-</div>
-
-__PerfSPEC__ has three phases:
-
-- Ranking
-- Learning
-- Runtime
-
-This repository is focused in __Learning__ phase with attention on:
-
-- Event logs info load and process
-- Predictive learning model 
-
-> Note: It is considered that __event data collection__ in `raw-audit-logs.log.gz` are realistic and representative to simulate
-administrative operations.
-
-## Files
-
-- `raw-audit-logs.log` contains raw Kubernetes audit logs collected using the `audit-policy.yaml` audit policy.
-
-Tools are distributed in directories:
-
-- [Collect](collect)
-- [Process](process)
-- [Learning](learning)
-
-As some tasks can be used in [Python](https://python.org) or [Rust](https://www.rust-lang.org/) there are directories for each programming languge inside directories tasks.  
-
-Each `task/programming-language` may have a __data__ directory where processing output files is generated. 
-
-### Collect data
-
-If you wish to [collect](collect) your own dataset, there are several source files that might help:
-
-- `collect/collect.py` is a script to trigger the installation and uninstallation of public Helm repositories.
-- `collect/helm-charts.json` is a backup of Helm charts used at the time of the collection.
-
-### Process data 
-
-
-### Learning
-
-
-
-## Reference
-
-[1]: [H. Kermabon-Bobinnec et al., "PerfSPEC: Performance Profiling-based Proactive Security Policy Enforcement for Containers," in IEEE Transactions on Dependable and Secure Computing, doi: 10.1109/TDSC.2024.3420712.](https://ieeexplore.ieee.org/document/10577533)

intro.md

@@ -1,57 +0,0 @@
-# PerfSPEC Learning Phase  INTRO
-
-Based in [PrefSPEC: Performance Profiling-based Proactive Security Policy Enforcement for Containers](https://ieeexplore.ieee.org/document/10577533) document presented in [1], thir repository contains source files used to generate and process data.
-
-[PrefSPEC document](PerfSPEC.pdf)
-
-[Presentación in Spanish](presentacion.pdf)
-
-<div style="margin: auto">
- <a target="_blank" href="perfspec-learning/src/branch/main/presentacion.pdf"><img src="imgs/prefSPEC-learning.png" width="800"></a>
-</div>
-
-__PerfSPEC__ has three phases:
-
-- Ranking
-- Learning
-- Runtime
-
-This repository is focused in __Learning__ phase with attention on:
-
-- Event logs info load and process
-- Predictive learning model 
-
-> Note: It is considered that __event data collection__ in `raw-audit-logs.log.gz` are realistic and representative to simulate
-administrative operations.
-
-## Files
-
-- `raw-audit-logs.log` contains raw Kubernetes audit logs collected using the `audit-policy.yaml` audit policy.
-
-Tools are distributed in directories:
-
-- [Collect](collect)
-- [Process](process)
-- [Learning](learning)
-
-As some tasks can be used in [Python](https://python.org) or [Rust](https://www.rust-lang.org/) there are directories for each programming languge inside directories tasks.  
-
-Each `task/programming-language` may have a __data__ directory where processing output files is generated. 
-
-### Collect data
-
-If you wish to [collect](collect) your own dataset, there are several source files that might help:
-
-- `collect/collect.py` is a script to trigger the installation and uninstallation of public Helm repositories.
-- `collect/helm-charts.json` is a backup of Helm charts used at the time of the collection.
-
-### Process data 
-
-
-### Learning
-
-
-
-## Reference
-
-[1]: [H. Kermabon-Bobinnec et al., "PerfSPEC: Performance Profiling-based Proactive Security Policy Enforcement for Containers," in IEEE Transactions on Dependable and Secure Computing, doi: 10.1109/TDSC.2024.3420712.](https://ieeexplore.ieee.org/document/10577533)

learning/python/lib_perfspec.py

@@ -1,6 +1,6 @@
 import marimo
 
-__generated_with = "0.10.17"
+__generated_with = "0.10.16"
 app = marimo.App(width="medium")
 
 
@@ -82,7 +82,7 @@ def perfspec_vars(app_vars, mo):
 
     perfspec = dict(
       defaults = dict(
-            epochs=300,
+            epochs=400,
             train_size=0.8,
             sequence_length=2,
             model_path="perfSPEC_model.keras",

learning/python/train_perfspec.py

@@ -1,15 +1,6 @@
-# /// script
-# requires-python = ">=3.13"
-# dependencies = [
-#     "keras==3.8.0",
-#     "marimo",
-#     "numpy==2.2.2",
-# ]
-# ///
-
 import marimo
 
-__generated_with = "0.10.17"
+__generated_with = "0.10.16"
 app = marimo.App(width="medium")
 
 
@@ -467,21 +458,15 @@ def f1_score_metric(PrecisionMetric, RecallMetric, keras, mo, tf):
 
 
 @app.cell(hide_code=True)
-def custom_validation_metrics(mo, tf):
+def custom_validation_metrics(X_val, mo, tf, y_val):
     #Custom callback to compute metrics on validation data
     class CustomValidationMetrics(tf.keras.callbacks.Callback):
-        def __init__(self, X_val, y_val):
-            super().__init__()  # Initialize the parent class
-            self.X_val = X_val
-            self.y_val = y_val
-
         def on_epoch_end(self, epoch, logs=None):
-            val_predictions = self.model.predict(self.X_val, verbose=0)
+            val_predictions = self.model.predict(X_val, verbose=0)
             val_predictions = (val_predictions > 0.5).astype(int)  # Binarize predictions
 
-            # Compute precision, recall, and f1-score
-            precision = tf.keras.metrics.Precision()(self.y_val, val_predictions)
-            recall = tf.keras.metrics.Recall()(self.y_val, val_predictions)
+            precision = tf.keras.metrics.Precision()(y_val, val_predictions)
+            recall = tf.keras.metrics.Recall()(y_val, val_predictions)
             f1_score = 2 * (precision * recall) / (precision + recall + 1e-7)
 
             print(f"\nEpoch {epoch + 1} Validation Metrics - Precision: {precision:.4f}, Recall: {recall:.4f}, F1 Score: {f1_score:.4f}")
@@ -558,7 +543,7 @@ def show_train_model_shape(mo, perfspec, prepare_train, verbose):
 
 
 @app.cell(hide_code=True)
-def make_model(CustomValidationMetrics, mo, np, perfspec):
+def make_model(mo, np, perfspec):
     #  Define the LSTM model
     def make_model(X=[],y=[],label_encoder=[], encoded_actions=[]):
         if len(X) == 0 or len(y) == 0:
@@ -584,13 +569,13 @@ def make_model(CustomValidationMetrics, mo, np, perfspec):
         perfspec['vars']['model'] = Sequential(
             [
                 #Embedding(input_dim=vocab_size, output_dim=embedding_dim),
-                Input(shape=(perfspec['settings']['sequence_length'], 1)),
                 LSTM(
                     perfspec['settings']['lstm_units_1'],
                     return_sequences=True,
                     recurrent_dropout=perfspec['settings']['dropout_rate'],
-                    input_shape=(perfspec['settings']['sequence_length'], 1),
-                ),            
+                    #input_shape = (2,vocab_size),
+                ),
+                Input(shape=(perfspec['settings']['sequence_length'], 1)),
                 LSTM(
                     perfspec['settings']['lstm_units_2'],
                     return_sequences=False,
@@ -637,12 +622,11 @@ def make_model(CustomValidationMetrics, mo, np, perfspec):
 
         # Callbacks
         early_stopping = EarlyStopping(
-            monitor="val_loss", patience=10, restore_best_weights=True
+            monitor="val_loss", patience=5, restore_best_weights=True
         )
         lr_reduction = ReduceLROnPlateau(
-            monitor="val_loss", patience=8, factor=0.8, min_lr=0.0001
+            monitor="val_loss", patience=3, factor=0.5, min_lr=0.0001
         )
-        custom_metrics_callback = CustomValidationMetrics(X, y)
         if perfspec['settings']['checkpoint_mode'] == "weights":
             # Save only the weights of the model instead of the full model.
             checkpoint = ModelCheckpoint(
@@ -661,9 +645,8 @@ def make_model(CustomValidationMetrics, mo, np, perfspec):
                 verbose=1                  # Print messages when saving
             )
 
-        callbacks=[early_stopping,lr_reduction]
-        callbacks=[early_stopping,lr_reduction]
-        callbacks.append(custom_metrics_callback)
+        callbacks=[early_stopping,lr_reduction]  #,CustomValidationMetrics] 
+        callbacks=[]  #,CustomValidationMetrics] 
         if checkpoint != None:
             callbacks.append(checkpoint)
 
@@ -693,7 +676,7 @@ def make_model(CustomValidationMetrics, mo, np, perfspec):
 
         This is where **model** is creates and **fit**
 
-        Saved in `perfspec['vars']` as `model` and `history`
+        Saved in `perfspec['vars'] as `model` and `history`
         """
     )
     return (make_model,)
@@ -808,11 +791,10 @@ def perfspec_save_model(Path, mo, perfspec):
 def perfspec_plot_history(Path, mo):
     def plot_history(perfspec):
         import json
-        from keras.src.callbacks import History
         if 'vars' not in perfspec:
             return None
         if perfspec['vars']['history'] != None:
-            if isinstance(perfspec['vars']['history'], History):
+            if 'history' in perfspec['vars']['history']:
                 _model_history = perfspec['vars']['history'].history
             else:
                 _model_history = perfspec['vars']['history']
@@ -1011,9 +993,8 @@ def perfspec_evaluate_model(Path, mo, np, prepare_train):
 
     def history_info(perfspec):
         import json
-        from keras.src.callbacks import History
         if perfspec['vars']['history'] != None:
-            if isinstance(perfspec['vars']['history'], History):
+            if 'history' in perfspec['vars']['history']:
                 model_history = perfspec['vars']['history'].history
             else:
                 model_history = perfspec['vars']['history']
@@ -1026,7 +1007,7 @@ def perfspec_evaluate_model(Path, mo, np, prepare_train):
                     model_history = json.load(history_file)
         if model_history != None:
             from prettytable import PrettyTable
-            train_loss = model_history['loss']
+            rain_loss = model_history['loss']
             val_loss = model_history['val_loss']
             train_acc = model_history['accuracy']
             val_acc = model_history['val_accuracy']