In many experiments with various cancer cells I found to get good results using: GTF2B. Note, this was comparing conditions with cancer cell lines (breast, cervix colon). But whereas other genes like: ACTB, HPRT, B2M,18S varied.... GTF2B remained unchanged.
I work in a system where it's very hard to find a good housekeeping gene that remains stable across conditions. We've begun using spike-in RNA controls. These controls are added to all samples during the extraction process and in the same amount. You then have an RNA that should be completely the same across conditions (skewed only due to errors in extraction). Here are a few papers that talk about spike-in controls.
Article An Internal Reference Technique for Accurately Quantifying S...
Article The use of endogenous and exogenous reference RNAs for quali...
Article Expression of prostaglandin D synthetase during development ...
Article Assessment of RT-qPCR Normalization Strategies for Accurate ...
when you say upregulated, do you mean with reference to the number of cells or to the amount of RNA? If the upregulation of housekeeping genes is due to cell death (RNA degradation) then it is a matter of lost RNA quality and there is no point of reference. RNA will degrade faster than rRNA and technically faster than housekeeping genes.