Home Vulnerability Critical Flaws in Ollama AI Framework Could Enable DoS, Model Theft, and Poisoning