CWP7 Resource Shield AI

AI-assisted server defense for CWP7 using Nebius Token Factory and NVIDIA Nemotron to explain incidents, assess severity, and guide safe defensive response.

Comment

Inspiration

CWP7 Resource Shield was created to solve a real hosting problem: one attacked WordPress or WooCommerce site can exhaust PHP-FPM, CPU and RAM and degrade an entire CWP7 server. The production product already detects coordinated abusive traffic and can push confirmed blocks to Cloudflare. For this hackathon, we are adding a new AI Security Analyst layer so operators can understand why an incident was flagged, what the evidence means, how severe it is, and what safe defensive action to take next.

What it does

CWP7 Resource Shield AI consumes structured, sanitized server-security telemetry and sends it to NVIDIA Nemotron through Nebius Token Factory. The analyst returns an operator-ready incident report with severity, confidence, attack-pattern explanation, evidence summary, affected assets, indicators, defensive recommendations, and a customer-safe summary.

The AI component is advisory and defensive. It does not autonomously scan third parties, exploit systems, retaliate, or execute arbitrary shell commands. Recommendations are constrained to reversible defensive actions such as edge blocking, rate limiting, log review, Cgroups checks, PHP-FPM review, WordPress hardening, credential review, backups, and continued monitoring.

How we built it

The hackathon edition is a standalone FastAPI application with a structured incident schema, Resource Shield telemetry adapter, Nebius Token Factory runtime integration, NVIDIA Nemotron reasoning, and a browser-based demonstration dashboard. The Token Factory API is used through its OpenAI-compatible endpoint. The app can discover available Nemotron models at runtime and supports explicit model selection for reproducible judging.

Significant update during the hackathon

The commercial CWP7 Resource Shield project existed before the submission period. During the hackathon, we significantly expanded it with a new AI Security Analyst architecture, Nebius Token Factory integration, NVIDIA Nemotron inference, sanitized telemetry adapter, standalone demo API/UI, and open-source hackathon edition. The proprietary commercial detection, licensing and production automation engine is not part of the MIT-licensed hackathon repository.

Why Nebius + NVIDIA Nemotron

The live Token Factory account exposes multiple NVIDIA Nemotron variants. We use Nemotron as the reasoning layer for evidence-aware cybersecurity incident interpretation while keeping deterministic detection in the existing security engine. This separation is important: rules and telemetry determine what happened; Nemotron helps a human operator understand what it means and what to do next.

Security and privacy

The demo avoids sending passwords, API keys, request bodies, WordPress content, customer data or raw authentication logs to the model. Only structured incident facts required for defensive analysis are transmitted.

Target track

Best Apps and Agents

Built With

  • cloudflare
  • cwp7
  • cybersecurity
  • fastapi
  • jinja
  • nebius-token-factory
  • nvidia-nemotron
  • openai-compatible-api
  • pydantic
  • python
Share this project:

Updates