In the case of the hair follicle staining, non-specific signal is inevitable as you can see the attached file.

Needless to say, when you perform blocking process in IHC, you have to use 3% BSA/ PBS instead of skim milk. 

The problem depends on the tissue quality or the primary antibody. Note that goat secondary antibody tends to show the non-specific signal. If there are any other primary antibodies with rabbit, rat or mouse, you have purchase and use them.  

http://dev.biologists.org/content/134/15/2795

Hi Iwan

You can try some troubleshooting with your procedure of IF. First, you need to check the clonal nature of your antibody. Since you are getting false positive, it would be better if you are using a monoclonal antibody. This will prevent chances of cross-reactivity of primary antibody with non-specific antigens. 

Secondarily, you can try choosing your blocking reagent more wisely. If you are targeting a phosphorylated protein for IF, you should use 3-5% BSA, whereas skimmed milk works good with non-phosphorylated proteins. You may also use a simple 3% blocking reagent and incubate your sample for some extended time, say 4 hours at room temperature. Following proper washings and incubations with primary and secondary antibodies, the results are very likely to be specific in nature. 

Good luck to you. 

Hi Iwan:

    Except for above constructive suggestions, proper controls are always the necessary and effective ways to tell false positive from real positive both in WB and IF.

    If possible, in your WB work, you can choose some your target protein overexpressed or depleted cells or tissue as the controls which can validate whether your assumed bands are real or not. For IF works, non-specific fluorescence is unavoidable, you always can get some clear images if you increase PMT, gains, laser power bla bla. To avoid misleading, non-staining, isotype controls, single staining slides are the necessary samples which can help you setup your IF parameters.

    An quick and easy way to figure out whether the signals from IF are real or not is by FACS. Going through your Ab stained samples in FACS machine comparing with your isotype control sample can give your primary idea whether the signal is real or not.