You need to be careful about using normalizers from online
Despite raising awareness of potential pitfalls of qPCR, use of experimentally validated reference genes for data normalisation is still widely disregarded
You should make a list of few miRNAs that are expressed in the serum, i.e. figure 6 of this article (always consult more than 1 article for a list)
Then you will need test whether the miRNAs from your candidate list are stable across your groups. Try to make a panel of >4 miRNAs.
Then you will need to use a normalization software such as GeNORM/Normfinder/bestkeeper (google them, your PCR software may be able to do it, i know BioRad's CFX software does it).
If you have found a miRNA paper which uses miRNA X as a normalizer, make sure that they have done this normalization process. The use of non-validated normalizers can result in frivolous results.