You generally should not normalize microarrays to housekeeping genes unless you have a very good reason to.

See full explanation here: http://www.people.vcu.edu/~mreimers/OGMDA/normalize.expression.html

One early approach was to select one gene, or a small subset of genes, which were not expected to be differentially expressed across any samples; these were usually chosen to be ‘housekeeping ’genes: genes which are required for basic cell processes, and which were once believed to have very stable expression levels in all cells. This commonsense approach is used routinely in rt-PCR. However, these housekeeping genes seem to vary by 30% or more across healthy samples, and even more in cancer samples (Lee et al, 2002). A 30% error in normalization is not a big deal for rt-PCR, where the accuracy is at best a half-cycle, and one cycle corresponds to an increase of a factor of two. However, in a microarray study a 30% difference over the whole genome can give many false positives.

See previous discussion here: Normalization based on housekeeping genes