It differs from most existing WAF, as it does not rely on signatures, but on a rather behavior analysis model.
This approach allows the WAF to be extremely lightweight, with nearly no performance impact ( and configuration, that are handled by native learning features embedded in the software.