🛡️

Advanced Malware Detection System

Powered by Machine Learning & Random Forest Algorithm

98.56%
Accuracy
134,435
Samples Analyzed
2,381
Features Extracted
581
Malware Families

Scroll down to explore more ↓

🎯 About This Project

This is a state-of-the-art malware detection system built using machine learning techniques. The model has been trained on the BODMAS dataset, containing over 134,000 real-world samples from August 2019 to September 2020. Using advanced feature extraction and Random Forest algorithm, we achieve industry-leading accuracy in detecting malicious software.

✨ Key Features

Machine Learning
🤖

Machine Learning Powered

Utilizes Random Forest, Gradient Boosting, and Logistic Regression algorithms for robust malware detection with 98.56% accuracy. Our advanced ML models learn from thousands of malware samples to identify threats effectively.

Analysis
📊

Comprehensive Analysis

Analyzes 2,381 features extracted using LIEF library to identify malicious patterns and behaviors in executable files. Deep feature extraction ensures thorough examination of potential threats.

Real-Time
🔍

Real-Time Detection

Test any file or dataset sample instantly. Get immediate results with confidence scores and detailed analysis. Fast processing without compromising system performance.

Performance
📈

High Performance

Achieves 98.31% true positive rate and 98.76% true negative rate, comparable to commercial antivirus solutions. Industry-leading accuracy with minimal false positives.

Feature Selection
🎯

Feature Selection

Intelligent feature selection reduces 2,381 features to top 200 most important ones for optimal performance. This optimization ensures faster processing while maintaining high accuracy.

Production
🛡️

Production Ready

Tested on real-world data with comprehensive evaluation metrics. Ready for deployment in security environments. Production-grade reliability and performance.

🛠️ Technology Stack

Python
Scikit-Learn
Random Forest
Gradient Boosting
Logistic Regression
LIEF Library
NumPy
Pandas
Matplotlib

🚀 Ready to Test?

Upload a file or enter an index number from the training dataset to see the model in action. Get instant predictions with detailed confidence scores and analysis.

💡 Dataset Information

📁 Dataset Name: BODMAS
📅 Time Period: Aug 2019 - Sep 2020
📊 Total Samples: 134,435
✅ Benign Files: 77,142
⚠️ Malware Files: 57,293
👥 Malware Families: 581
🔧 Features: 2,381 (LIEF extracted)
🎯 Selected Features: 200 (top importance)
📈 Train/Test Split: 80/20

📟 Terminal/Console Output

▶ python malware_detection.py
======================================================================
MALWARE DETECTION PROJECT - BODMAS DATASET
======================================================================

📚 Dataset Info:
- Total Samples: 134,435 (57,293 malware + 77,142 benign)
- Features: 2,381 (extracted using LIEF)
- Time Period: August 2019 - September 2020
- Malware Families: 581

======================================================================

📥 STEP 1: Data Loading
----------------------------------------------------------------------
🔄 Loading feature vectors...
✅ Data loaded successfully!
Features shape: (134435, 2381)
Labels shape: (134435,)
Benign samples: 77142
Malware samples: 57293

📊 STEP 2: Exploratory Data Analysis
----------------------------------------------------------------------
🔄 Calculating feature importance...
✅ EDA visualization saved as 'eda_analysis.png'

🔧 STEP 3: Feature Selection & Preprocessing
----------------------------------------------------------------------
✅ Selected top 200 features
Original features: 2381
Selected features: 200
✅ Data normalization complete

🔀 STEP 4: Train-Test Split
----------------------------------------------------------------------
✅ Data split complete:
Training samples: 107548 (80.0%)
Testing samples: 26887 (20.0%)
Train - Benign: 61714, Malware: 45834
Test - Benign: 15428, Malware: 11459

🤖 STEP 5: Model Training
----------------------------------------------------------------------
🔄 Training Random Forest...
Cross-validation accuracy: 0.9847 (+/- 0.0012)
✅ Random Forest training complete

🔄 Training Gradient Boosting...
Cross-validation accuracy: 0.9782 (+/- 0.0015)
✅ Gradient Boosting training complete

🔄 Training Logistic Regression...
Cross-validation accuracy: 0.9623 (+/- 0.0018)
✅ Logistic Regression training complete

📊 STEP 6: Model Evaluation
----------------------------------------------------------------------
======================================================================
RANDOM FOREST
======================================================================

✨ Accuracy: 0.9856 (98.56%)
✨ ROC-AUC Score: 0.9912

📋 Classification Report:
precision recall f1-score support
Benign 0.9889 0.9876 0.9882 15428
Malware 0.9812 0.9831 0.9821 11459
accuracy 0.9856 26887

📊 Confusion Matrix:
True Negatives (Benign correctly): 15236
False Positives (Benign as Malware): 192
False Negatives (Malware as Benign): 194
True Positives (Malware correctly): 11265

======================================================================
🏆 Best Model: Random Forest
Accuracy: 0.9856
ROC-AUC: 0.9912
======================================================================

✅ Project complete! Check the generated visualizations.

📊 Detailed Statistics

Total Samples

134,435

Real-world malware dataset

Benign Files

77,142

Safe software samples

Malware Files

57,293

581 malware families

Features

2,381

Extracted using LIEF

📈 Training Progress

Random Forest Training

98.5% Accuracy

Gradient Boosting Training

97.8% Accuracy

Logistic Regression Training

96.2% Accuracy

🎯 Performance Metrics

🌲 Random Forest

98.56%

Best Overall Performance

⭐ Recommended

📈 Gradient Boosting

97.82%

Good Balance

📊 Logistic Regression

96.23%

Fast & Efficient

📸 Model Results Visualization

Explore our model's performance through detailed visualizations and metrics.

Model Performance Metrics

Model Performance Metrics

This visualization shows the comprehensive performance metrics of our Random Forest model, including accuracy, precision, recall, and F1-score across different malware categories. The model demonstrates exceptional performance with 98.56% accuracy.

📊 Model Performance Highlights

🎯
98.56%
Overall Accuracy
📈
0.9912
ROC-AUC Score
✅
98.31%
Malware Detection Rate
🛡️
98.76%
Benign Identification
⚡
1.24%
False Positive Rate
🔒
1.69%
False Negative Rate

🏆 Final Results Summary

Best Model: Random Forest

98.56%

Overall Accuracy

📋 Detailed Performance Breakdown

Metric Random Forest Gradient Boosting Logistic Regression
Accuracy 98.56% 97.82% 96.23%
ROC-AUC 0.9912 0.9856 0.9734
Precision (Benign) 98.89% 97.65% 96.12%
Recall (Malware) 98.31% 97.23% 95.87%
F1-Score 98.52% 97.78% 96.19%
Training Time ~5 min ~8 min ~2 min

🎯 Confusion Matrix (Random Forest)

Predicted
Benign Malware
Actual Benign 15,236 192
Malware 194 11,265

✅ Key Achievements:

  • True Positive Rate: 98.31% (11,265 out of 11,459 malware detected)
  • True Negative Rate: 98.76% (15,236 out of 15,428 benign files identified)
  • False Positive Rate: Only 1.24% (192 benign files misclassified)
  • False Negative Rate: Only 1.69% (194 malware files missed)

💡 What This Means:

Out of every 100 files:
• 98-99 will be correctly classified
• Only 1-2 will be misclassified
• This is production-ready performance!
• Comparable to commercial antivirus solutions

🔍 Test Your Model

Provide an index (0–134,434) from bodmas.npz. Index 0 is BENIGN by design, while index 57,293 is the first MALWARE entry—use these for live demos.

📤 Upload File for Prediction

Upload a Windows executable file (.exe, .dll) to analyze it for malware.
OR

📊 Dataset Lookup

Enter an index from the BODMAS dataset to test with a known sample.
Loading...
Prediction Result
0%
📊 Prediction: -
🎯 Confidence: -
📈 Model Accuracy: 98.56%
🤖 Model Used: Random Forest
⏱️ Processing Time: -

📋 Additional Information:

-