AyeshaaRafaqat/Prompt-Injection-Detection-Engine

# Prompt Injection Detection Engine (PIDE) A multi‑layer detection system for prompt‑injection attacks against large language models. It provides a fast, rule‑based Layer 1, embedding similarity, heuristic analysis and risk scoring. ## Quick Start (no extra files needed) ### 1. Clone the repository git clone https://github.com/AyeshaaRafaqt/Prompt-Injection-Detection-Engine.git cd Prompt-Injection-Detection-Engine ### 2. Choose how to run it #### a) **Using the bundled Makefile** (recommended for beginners) # Create virtual environment and install dependencies make install # Start the API server (FastAPI) – will be reachable at http://localhost:8000 make run-api # In another terminal, launch the Gradio playground UI – http://localhost:7860 make run-ui # Run the full test suite make test #### b) **Manual commands** (if you prefer the raw steps) # Windows PowerShell python -m venv venv .\venv\Scripts\activate pip install -r requirements.txt python -m spacy download en_core_web_sm python scripts/build_exemplars.py # API uvicorn api.main:app --host 0.0.0.0 --port 8000 --reload # Gradio UI python -m demo.gradio_app # Run Tests pytest tests/ -v #### c) **Docker** (one‑liner for any platform with Docker installed) # Build the image (once) docker build -t pide . # Run the container – exposes API on 8000 and UI on 7860 docker run -p 8000:8000 -p 7860:7860 pide ## Project Layout (for reference) . ├── api/ # FastAPI REST gateway ├── config/ # YAML patterns & scoring config ├── data/ # Exemplars & FAISS index ├── demo/ # Gradio UI & LLM client helper ├── evaluation/ # Ablation & benchmark scripts ├── layers/ # Detection layers (L1‑L4) ├── logs/ # Audit logs (privacy‑preserving) ├── scripts/ # Utility scripts (e.g., build_exemplars.py) ├── tests/ # Pytest suite (39 tests) ├── Dockerfile # Container build file ├── Makefile # Helper targets for common tasks ├── pipeline.py # Orchestrates layers & fail‑secure logic ├── requirements.txt └── README.md # (this file) ## Dockerfile (included in repo) The repository already contains a minimal `Dockerfile` that: 1. Uses a lightweight Python 3.11 base image. 2. Creates a virtual environment, installs all dependencies, and downloads the spaCy model. 3. Exposes ports **8000** (API) and **7860** (Gradio UI). 4. Starts the API server by default. The UI can be accessed via the same container at `http://localhost:7860`. ## Makefile (included in repo) VENV=venv PYTHON=$(VENV)/Scripts/python.exe # Create virtual environment venv: python -m venv $(VENV) # Install dependencies and spaCy model install: venv $(PYTHON) -m pip install --upgrade pip $(PYTHON) -m pip install -r requirements.txt $(PYTHON) -m spacy download en_core_web_sm # Run API server (FastAPI) run-api: $(PYTHON) -m uvicorn api.main:app --host 0.0.0.0 --port 8000 --reload # Run Gradio Playground UI run-ui: $(PYTHON) -m demo.gradio_app # Execute the full test suite test: $(PYTHON) -m pytest -q Run any target with `make ` (e.g., `make run-ui`). ## License MIT – feel free to fork, extend, and use in your own projects.