Contact person: Business Client

Phone:Show

Email:Show

Location: Miami, United States

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I need an AI-driven protection layer for a network of IoT devices that have been experiencing persistent electronic harassment. The solution should learn the normal traffic patterns of each device and immediately block or quarantine anything anomalous. Because our companion control app runs on Android, the model also needs a clean API that the app can call for real-time threat reports and rule updates.

You are free to work in TensorFlow, PyTorch, scikit-learn, or another lightweight embedded framework, provided training remains reproducible and inference can run on a modest ARM-based gateway. I will supply device firmware specs, traffic logs, and a full test sandbox once we get started.

Deliverables
• Trained model with documented architecture
• Inference script or compiled module ready to embed on the IoT gateway (≤ 256 MB RAM)
• REST or MQTT interface compatible with the Android app
• Setup guide covering data prep, training pipeline, and deployment steps
• Brief report on detection accuracy and false-positive rate

Acceptance criteria
The model must detect at least 95 % of hostile packets in our sample set while keeping false positives below 3 %, and it must load and run on the target hardware without external dependencies beyond the provided runtime." (client-provided description)