To be sure, whether it is unspecific antibody binding, try normal rabbit serum detected with your secondary antibody. If you get the same signal, you can be sure it is unspecific.

In order to block unspecific binding you might try to preincubate your blot with normal donkey serum; a concentration of 5% at minimal volume is usual.

Finally, add 0,1% tween 20 to your incubation buffer to reduce unspecific binding. At that concentration it doesn't influence specific binding too much. If it does, you had very week binding, anyway.

[Does it mean that the strong band on the blot doesn't come from the protein of interest but it is non-specific binding] In principle you can never totally exclude cross-reactivities of any antibody even the wonderful monoclonals, so what is unspecific is often pretty difficult to say. This depends on the antigen in question and the tissues studied, and in most cases on the primary antibody or antiserum. You can overcome this task only by careful and sophisticated analyses and controls. It often helps to try to "dilute out" the "light/weak bands" by diluting the primary antibody, but there is no guarantee to that method, another time it helps trying various blocking reagents - and this is a whole science in itself. Thus, blot preincubations in a normal serum from the same animal as the one producing the secondary detection antibody (here donkey) and even keeping this additive in all antibody solutions (1-5%) is a good idea together with some detergents (0.1% Tween 20 or 80) or somewhat higher salt conditions (1.8% NaCl) or even some sugars (most antibodies are glycosylated) may help keeping unspecific bindings down, unfortunately thsi is often very much a matter of trial and error. Leaving out the primary antibody or preabsorbing it with the respective antigen from the immunization regime prior to the Westen blot gives good answers about unspecifics, although it cannot necessarily exclude extra bands but helps identifying those bands that are definitely specific! Just proper controls (even incl. positive controls with the respective protein -if available even in crude form- run in parallel on the same PAGE) are the keys, as always... Good luck.

Thank you very much for your suggestions, I will try to find the best solution.