How to calculate or estimate robustness of an reversible data hiding algorithm?

I have implemented an image based RDH algorithm using histogram shifting. I am getting a decent embedding capacity and good visual quality marked images. 

As I know the three main characteristics of an RDH technique is embedding capacity, visual image quality and security or robustness. I understood the first two parameters well, but I am facing little problem in justifying robustness of my technique.

Please help me by providing some information.

Thank you.