Cloudflare modified its web application firewall (WAF) to add protections for applications that use large language models.
The service, called “Firewall for AI”, is available to enterprise customers of the Application Security Advanced enterprise. Προς το παρόν συμπεριλαμβάνει δύο δυνατότητες: Advanced Rate Limiting, και Sensitive Data Detection.
Advanced Rate Limiting allows the customer to create a policy that sets a maximum rate of requests made by a single IP address or API key during a period connections. Doing so helps prevent distributed denial of service (DDoS) attacks when using the model or other situations that would flood the LLM with requests and disrupt its ability to process legitimate queries.
The second feature, Sensitive Data Detection, prevents LLMs from leaking confidential data in query responses. It also allows customers to define WAF rules that scan for financial information such as credit card numbers and secrets such as API keys, to ensure that these sensitive details do not end up in an LLM's responses.
Unfortunately, there is still no firewall rule to prevent chatty models from spewing rubbish or bad and fabricated information.
In the future, customers will be able to “build their own custom fingerprints footprints (fingerprints)” and adjust what information the models may or may not reveal, according to Daniele Molteni, Cloudflare group product manager, who he said the Firewall for AI on Monday.
