CyberSentinel AI

CyberSentinel AI is a Kivy/KivyMD mobile application that acts as an AI-powered cybersecurity assistant. It supports online and offline Q&A, scheduled vulnerability scans (simulated for mobile), and persistent history with SQLite. The project is production-ready for Android packaging via Buildozer.

Features

• Modern cyber-themed UI using KivyMD (dark mode, neon accents)
• Chatbot with a friendly professional "cyber" personality
• Offline AI Q&A powered by a lightweight local knowledge base (and optional GPT4All)
• Online AI support: OpenAI and HuggingFace Inference (easily extendable to IBM Watson / Microsoft Copilot)
• Persistent SQLite memory of chats and scan logs
• Unanswered queue stored offline; auto-answers once online
• Periodic device scan (simulated): detects and isolates threats, notifies the user
• User controls for scan schedule, approvals, and API keys
• Animated avatar and cyber grid background

Project Structure

.
├── app
│   ├── backend
│   │   ├── ai_providers.py
│   │   ├── connectivity.py
│   │   ├── database.py
│   │   ├── persona.py
│   │   ├── scanner.py
│   │   └── scheduler.py
│   ├── main.py
│   └── ui
│       └── cybersentinel.kv
├── assets
│   ├── avatars
│   │   └── avatar.svg
│   └── icons
│       └── cybersentinel_icon.svg
├── data
│   └── local_knowledge_base.json
├── scripts
│   └── setup_db.py
├── buildozer.spec
├── requirements.txt
└── README.md

Local Development

1) Create a virtual environment

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip wheel setuptools
pip install -r requirements.txt

2) Initialize the database (optional)

python scripts/setup_db.py

The app will auto-create tables on first run as well.

3) Run the app on desktop

python app/main.py

Notes:

• The UI, scanning, and storage all work on desktop for development.
• Offline answers use the local knowledge base. For a local LLM, optionally install gpt4all and place models according to its documentation.

API Configuration

Open the Settings tab to set any available API keys. The app can use:

• OpenAI API (OPENAI_API_KEY)
• HuggingFace Inference API (HUGGINGFACE_API_KEY)

The app is architected to be extended to IBM Watson (watsonx) and Microsoft Copilot endpoints. You can add providers in app/backend/ai_providers.py.

Android Build (Buildozer)

Ensure you have the Android SDK/NDK and Buildozer installed. On Ubuntu:

sudo apt update
sudo apt install -y python3-pip build-essential git zip unzip openjdk-17-jdk
pipx install buildozer
pipx ensurepath
# Or install via pip if preferred:
# pip install --user buildozer

Then build the APK:

cd /workspace
buildozer android debug

The first build can take a while. The output APK will be under bin/.

Buildozer Spec Notes

• Requirements: python3,kivy,kivymd,requests,schedule,psutil,pillow,numpy
• Permissions: INTERNET, READ/WRITE EXTERNAL STORAGE (for demo isolation log)
• Orientation: portrait
• KV files are included automatically

If you plan to use GPT4All offline models on Android, you will need to:

• Add gpt4all and its dependencies to the buildozer requirements
• Bundle and load a small model that runs on-device (ensure size/performance fit)
• Adjust any necessary JNI/NDK options

Security Disclaimer

This app simulates scanning and isolation for demonstration only. It does not perform invasive or privileged operations. Do not rely on it for real endpoint protection.

Troubleshooting

• If the app fails to build on Android, try clearing the build cache:

buildozer android clean

• Ensure Java 17+ and recent SDK/NDK
• If OpenAI/HuggingFace requests time out, verify internet connectivity
• On desktop, window resizing may affect the animated grid; it is cosmetic only

License

MIT