PlagShield — Multi-Language Code Plagiarism Detection System

An AI-powered multi-language code plagiarism detection and visualization platform that combines token-based similarity analysis, AST structural comparison, semantic embeddings, and interactive visual analytics to identify copied, modified, or obfuscated source code submissions.

---

About The Project

PlagShield is an intelligent plagiarism detection platform designed to detect source code similarity across multiple programming languages using both traditional and AI-powered approaches.

Traditional plagiarism detection systems primarily rely on token matching techniques and often fail against advanced obfuscation strategies such as:

• Variable renaming
• Code restructuring
• Loop and control-flow transformations
• Cross-language translation
• Semantic modifications

PlagShield addresses these limitations by integrating:

• JPlag token-based similarity analysis
• AST (Abstract Syntax Tree) structural comparison
• Cross-language normalization
• CodeBERT semantic embeddings
• Machine learning-based plagiarism risk scoring
• Interactive visualization dashboards

The platform supports multiple programming languages including:

• Java
• Python
• C/C++
• JavaScript

PlagShield is designed for:

• Academic institutions
• Competitive programming platforms
• Technical assessment systems
• Software quality analysis
• Research environments

---

Features

• Multi-language plagiarism detection
• Token-based similarity analysis
• AST-based structural comparison
• Cross-language code normalization
• AI semantic similarity detection using CodeBERT
• Machine learning-based risk scoring
• Anomaly and collusion detection
• Interactive visualization dashboard
• Similarity heatmaps and network graphs
• Side-by-side code diff viewer
• Exportable PDF and CSV reports
• Batch processing support
• Modular multi-service architecture

---

System Architecture

Input Code Files
        ↓
Preprocessing Module
        ↓
Token-Based Analysis (JPlag)
        ↓
AST Structural Analysis
        ↓
Cross-Language Normalization
        ↓
AI Semantic Embeddings (CodeBERT)
        ↓
ML Risk Classification
        ↓
Clustering & Anomaly Detection
        ↓
Visualization Dashboard & Reports

---

Modules

1. Input Module

Handles batch upload and language identification of source code submissions.

2. Preprocessing Module

Performs:

• Comment removal
• Whitespace normalization
• Identifier anonymization
• Constant normalization

3. Token Similarity Module

Uses JPlag’s Greedy String Tiling (GST) algorithm for token-based similarity detection.

4. AST Structural Analysis Module

Analyzes Abstract Syntax Trees to detect structural similarities and obfuscation techniques.

5. Cross-Language Normalization Module

Maps language-specific constructs into a unified intermediate representation.

6. AI Embedding Module

Uses CodeBERT embeddings for semantic similarity analysis between code fragments.

7. ML Classifier & Risk Scoring Module

Combines all similarity features into a unified plagiarism risk score.

8. Clustering Module

Detects collusion groups and suspicious submission clusters.

9. Visualization Dashboard

Provides:

• Similarity heatmaps
• Network graphs
• Risk-ranked tables
• Code diff viewers

10. Report Generation Module

Generates:

• PDF reports
• CSV exports
• Statistical summaries

---

Tech Stack

Frontend

• React.js
• D3.js
• HTML
• CSS
• JavaScript

Backend

• Python
• Flask / FastAPI

AI & Machine Learning

• CodeBERT
• Transformers
• Scikit-learn

Database

• MongoDB / MySQL

Tools & Frameworks

• JPlag
• Git & GitHub

---

Installation

Prerequisites

Make sure the following tools are installed on your system:

• Git
• Python 3.10 or newer
• Node.js 20 or newer (npm included)
• Windows PowerShell

Optional (Only for Spring Backend)

• Java 17
• Maven 3.9+

---

1. Clone the Repository

git clone https://github.com/Shivank2005/Code_Plagarism.git
cd Code_Plagarism

---

2. Create and Activate a Python Virtual Environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1

If PowerShell Blocks Activation

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass

---

3. Install Python Dependencies

pip install -r requirements.txt
pip install flask flask-cors

---

4. Install Frontend Dependencies

cd frontend
npm install
cd ..

---

5. Start the Full Application Stack

.\start-all.ps1

This starts:

• CodeBERT Service → Port 8090
• Backend API → Port 8082
• Frontend → Port 5173

Open in browser:

http://localhost:5173/

---

Manual Start (Alternative)

If you prefer running services separately:

Terminal 1 — CodeBERT Service

cd codebert-service
python app.py

---

Terminal 2 — Backend API

cd backend
python mock_app.py

---

Terminal 3 — Frontend

cd frontend
npm run dev

---

Quick Health Checks

Frontend

http://localhost:5173/

Backend Health

http://localhost:8082/health

CodeBERT Service

http://localhost:8090/

---

Usage

1. Upload source code submissions
2. Select supported programming languages
3. Run plagiarism analysis
4. View similarity scores and visual reports
5. Export plagiarism reports

---

Visualization Features

The dashboard includes:

• Similarity Heatmaps
• Submission Network Graphs
• Risk-Ranked Submission Lists
• Cluster Visualization
• Side-by-Side Code Diff Viewer

---

---

Contributors

• Shivank

---

License

This project is developed for academic and research purposes.

---

References

• JPlag
• CodeBERT
• GraphCodeBERT
• CodeXGLUE
• AST-based clone detection research papers

---