I think the key to comprehending this passage is the first sentence of the paragraph it is in, which refers to a "random oracle paradigm". From wikipedia, a random oracle is essentially a black-box function; deterministic and repeatable, yet its output is indistinguishable from random noise. The random oracle paradigm is defining the security of a cryptographic primitive by comparing it to a random oracle, i.e. does it leak any information about its function such that its output can be distinguished from random noise. This, the paper argues, is a reasonable standard for symetric primitives, but unreasonable for asymmetric ones. The crux of their argument is, in essence, secure symmetric primitives have very complex internal function, but asymmetric primitives are comparable very simple. Thus, as the sentence you provide very unclearly states, it is unreasonable to think that an attacker would have no knowledge of the internal function of asymmetric primitives, and so it is not necessary that an asymetric primitive's output leak no information about its function, i.e. indistinguishable from random noise, for it to be secure.