🛡️ PromptShield

AI-Powered LLM Firewall for Prompt Injection Detection & Secure AI Applications

PromptShield is an enterprise-grade AI security middleware that protects Large Language Models (LLMs) from malicious prompts such as Prompt Injection, Jailbreak Attacks, Role Manipulation, and PII Extraction attempts before they ever reach the model.

Built using FastAPI, React, Hugging Face Transformers, Microsoft Presidio, Docker, and Google Gemini, PromptShield provides real-time detection, explainable analytics, and an interactive security dashboard.

---

🚀 Features

🔐 Multi-Layer AI Security

• ✅ Prompt Injection Detection
• ✅ Jailbreak Detection
• ✅ Role Manipulation Detection
• ✅ PII Extraction Detection
• ✅ Risk Scoring Engine
• ✅ Allow / Block Decision Engine

---

🤖 AI & NLP

• Fine-tuned DistilBERT classifier
• Microsoft Presidio for PII detection
• Custom Indian PII recognizers
• Rule-based jailbreak detection
• Google Gemini integration

---

📊 Security Dashboard

• Live Threat Monitoring
• Threat Trend Analytics
• Attack Distribution Charts
• Detection Module Statistics
• Threat Intelligence Cards
• Recent Attack Feed
• Incident Details Panel
• CSV Export
• Advanced Filters

---

🐳 Production Deployment

• Dockerized Backend
• Dockerized Frontend
• Docker Compose
• Health Checks
• Persistent SQLite Database
• Production-ready Containers

---

📸 Dashboard Preview

(Screenshots coming soon)

---

🏗 Architecture

                    User Prompt
                         │
                         ▼
              PromptShield Middleware
                         │
        ┌────────────────┼────────────────┐
        │                │                │
        ▼                ▼                ▼
 Prompt Injection    PII Detection   Jailbreak Detection
    Detection          (Presidio)        Rule Engine
        │                │                │
        └────────────────┼────────────────┘
                         ▼
                 Risk Scoring Engine
                         │
               Allow / Block Decision
                         │
                         ▼
                  Target LLM (Gemini)
                         │
                         ▼
                  Security Dashboard

---

⚙ Tech Stack

Backend

• Python
• FastAPI
• Hugging Face Transformers
• PyTorch
• Microsoft Presidio
• SQLite
• Google Gemini API

Frontend

• React
• Vite
• Axios
• Recharts
• Framer Motion

Deployment

• Docker
• Docker Compose
• Nginx

---

📂 Project Structure

PromptShield
│
├── backend
│   ├── app
│   ├── trained_models
│   ├── promptshield.db
│   ├── Dockerfile
│   └── requirements-prod.txt
│
├── frontend
│   ├── src
│   ├── public
│   ├── Dockerfile
│   └── package.json
│
├── docker-compose.yml
│
└── README.md

---

🚀 Getting Started

Clone

git clone https://github.com/Dhimahi-T-Mehta/PromptShield.git

cd PromptShield

---

Backend

cd backend

pip install -r requirements-prod.txt

uvicorn app.main:app --reload

---

Frontend

cd frontend

npm install

npm run dev

---

🐳 Run with Docker

docker compose up --build

Backend

http://localhost:8000

Swagger

http://localhost:8000/docs

Frontend

http://localhost:5173

---

📊 Implemented Modules

• Prompt Injection Detection
• Jailbreak Detection
• Role Manipulation Detection
• PII Detection
• Risk Engine
• Security Pipeline
• Logging
• Dashboard Analytics
• CSV Export
• Explainable AI Dashboard
• Gemini Integration
• Docker Deployment

---

🎯 Current Status

✅ Phase 1 — Project Setup

✅ Phase 2 — Dataset Collection

✅ Phase 3 — Model Training

✅ Phase 4 — Detection Engine

✅ Phase 5 — FastAPI Backend

✅ Phase 6 — React Dashboard

✅ Phase 7 — Premium Cybersecurity UI

✅ Phase 8 — Security Modules

✅ Phase 9 — Explainable AI

✅ Phase 10 — Advanced Dashboard

✅ Phase 11 — LLM Integration

✅ Phase 12 — Docker & Production Deployment

🚧 Phase 13 — Redis Caching (In Progress)

---

🎓 Motivation

PromptShield was developed as a final-year Information Technology project to address one of the most critical security challenges in modern AI systems—Prompt Injection, ranked among the top risks in the OWASP Top 10 for LLM Applications.

The project demonstrates how AI-powered middleware can protect LLMs before malicious prompts reach the model.

---

👨‍💻 Author

Dhimahi Mehta

Final Year B.E. Information Technology Student

AI • Machine Learning • Cybersecurity

GitHub: https://github.com/Dhimahi-T-Mehta

---

⭐ If you found this project interesting, consider giving it a star!