Comprehensive protection for your AI applications
ZyCDN AI protection module targets LLM API gateways, providing prompt injection detection, sensitive content filtering, rate limiting, token usage statistics, and dynamic request encryption. Supports OpenAI, Claude, ERNIE Bot, Tongyi Qianwen, and other major LLM services.
Auto-detect and mask phone numbers, ID numbers, bank cards, and other sensitive data to prevent privacy leakage through AI APIs. Supports custom masking rules.
Prevent prompt injection attacks and jailbreak attempts. Protect system prompts from theft or tampering. Real-time detection and blocking of malicious prompt requests.
Fine-grained request rate control at user, IP, and API levels. Prevents abuse and malicious calls with dynamic threshold adjustment.
Intelligent caching based on semantic similarity. Auto-identifies semantically similar requests and returns cached results, reducing LLM API costs and latency.
Real-time token consumption monitoring with usage trend analysis, budget alerts, and cost reports for precise LLM API cost control.
Technical details about ZyCDN self-hosted CDN AI LLM protection.
ZyCDN AI protection module deploys on CDN edge nodes as an AI application gateway. When requests arrive, it performs: 1) Data masking to identify and hide sensitive info; 2) Prompt injection detection to block malicious injection and jailbreak attempts; 3) Rate limiting to control request frequency; 4) Semantic caching for similar requests; 5) Token usage statistics. All processing happens at the edge without adding origin pressure.
ZyCDN AI protection is model-agnostic, working via HTTP/HTTPS proxy. It supports all API-based LLM services including OpenAI GPT, Claude, ERNIE Bot, Tongyi Qianwen, ChatGLM, DeepSeek, and more. Simply point your AI application API requests to ZyCDN edge nodes to get masking, protection, rate limiting, caching, and statistics.
