🔒 SmartShield

SmartShield is an advanced malware detection and secure image-sharing system that leverages machine learning, steganography, browser-based security protocols, and TypeScript. It detects image-based malware and ensures that sensitive image files can only be accessed by the intended recipient or device.

---

🚀 Features

🛡️ Malware Detection

• Detects malware hidden in image files using an ensemble ML model.
• Uses ResNet50 for feature extraction and SVM, Random Forest, and Logistic Regression for classification.
• Integrated into a browser extension for real-time malware scanning.

🔐 Secure Image Sharing

• Steganographically embeds user/device metadata in images.
• Ensures only authorized users/devices can decode or access the image.
• Unauthorized access results in image corruption or destruction.

🌐 Browser Extension

• “Check Image” button for analyzing images before download.
• Option to manually upload and check images for malware threats.

---

🧠 Tech Stack

💻 Frontend

• HTML, CSS, JavaScript, TypeScript
• React (for extension UI)
• Tailwind CSS (for styling)

🔙 Backend

• Flask (Python)
• SQLAlchemy + SQLite
• Steganography libraries (cv2, Pillow)

🧠 AI/ML Tech Stack

• ResNet50 – Used as a feature extractor for malware images.
• Ensemble Classifier – Combination of:
  • Support Vector Machine (SVM)
  • Random Forest
  • Logistic Regression
• VotingClassifier – Implements soft voting for better accuracy.
• Malimg Dataset – Dataset of malware images used for training and evaluation.
• Joblib – Used for saving and loading the trained model (svm_model.pkl).

---

🖥️ Web Tech Stack

Layer	Tech Used
Frontend	React.js, Tailwind CSS
Backend	Flask (Python)
Steganography	Stegano (Python)
Database	MongoDB
Auth	Email/Password Login

---

🧪 How It Works

1. 🖼️ User uploads an image
2. 🔍 Image is analyzed using ResNet + ML ensemble
3. ✅ Safe images are either:
   • Downloaded normally, or
   • Embedded with user/device metadata
4. 🔓 Only the intended user/device can view or decode the image
5. 🚫 Unauthorized attempts lead to image corruption or destruction

---

🧠 Malware Detection Flow

• ResNet50 pre-trained model is used for feature extraction.
• Extracted features are passed to a soft voting ensemble model:
  • SVM (linear kernel)
  • Random Forest (n=100)
  • Logistic Regression
• Predicts one of the 25 malware classes from Malimg dataset.
• If confidence < threshold (e.g., 60%), returns "No Malware".

🔮 Future Scope

1. Device-Level Locking
   Bind image access to device fingerprints or MAC addresses. Ensures only specific machines can decode or view sensitive files.

2. Time-Based Access Control
   Auto-expire access after a set time window (e.g., 24 hours after download link is generated).

3. IP Whitelisting
   Restrict file access to predefined IP ranges (great for enterprise/internal networks).

4. Cryptographic Enhancement
   Integrate AES encryption in addition to steganography for dual-layer security.

5. Blockchain Audit Trails
   Maintain tamper-proof logs of image modifications, access attempts, and transmission.

6. Watermarking + Steganography
   Combine visible and invisible markings to prevent both unauthorized use and tampering.

---

🤖 LLM (Large Language Model) Integration

We plan to integrate an LLM-powered chatbot into the platform to assist with:

1. User Queries

   • "Why was my file blocked?"
   • "How can I securely send this image?"

2. Forensic & Security Insights

   • "This image was modified twice and attempted to be accessed by an unauthorized system."
   • "Tampering suspected. Recommend isolating source device."

3. Dynamic Policy Generation

   • “For healthcare images, only allow access from these 3 IPs and auto-delete after 24 hours.”

4. Security Recommendations

   • Based on image metadata and usage patterns, recommend firewall rules or access restrictions.

✅ Perfect for enterprise clients, security analysts, and forensics teams looking to enforce adaptive policies.

---

Project Strcuture

smartshield/ │ ├── backend/ │ ├── app.py │ ├── model/ │ │ └── ensemble_model.pkl │ └── utils/ │ └── steganography.py │ ├── frontend/ │ ├── index.html │ └── extension/ │ ├── manifest.json │ ├── popup.tsx │ ├── background.ts │ ├── utils/ │ │ └── imageUtils.ts │ └── style.css │ └── README.md

🧰 Installation

Backend (Flask API)

git clone https://github.com/yourusername/smartshield.git
cd smartshield/backend
pip install -r requirements.txt
python app.py