Many of the whitepapers are somewhat out-of-date with the implementation or too vague to serve as actual references for many of the algorithms being used so I am not sure about their usefulness as references. There have been newer papers about the high-level security features and resistance to attacks (like this one, and look at anything by Greig Paul from University of Strathclyde) but they do not provide concrete details about the core algorithms, especially about the distributed system part.
It would actually be super useful if we collectively extracted the algorithms and wrote other papers about them with more complete evaluation sections to serve as references.
What is the context of the paper you are writing? Where do you plan to submit it?
If you simply are looking for references that are associated with MaidSafe I would go for the newer papers. Otherwise, you might simply as well explain the part you need for your paper, give credit to MaidSafe for the contribution in the acknowledgement section, and build the rest of the paper around it.