Different probabilities for SVC in LIBSVM and sklearn 

I'm studying LIBSVM library and comparing it with sklearn implementation of SVM. I found that the probabilities obtained in SVC for sklearn and LIBSVM are different from each other. What can this be related to? 