Evidence combination based on CSP modeling

The evidence theory and its variants are mathematical formalisms used to represent uncertain as well as ambiguous data. The evidence combination rules proposed in these formalisms agree with Bayesian probability calculus in special cases but not in general. To get more reconcilement between the belief functions theory with the Bayesian probability calculus, this work proposes a new way of combining beliefs to estimate combined evidence. This approach is based on the Constraint Satisfaction Problem modeling. Thereafter, we combine all solutions of these constraint problems using Dempster's rule. This mathematical formalism is tested using information system security risk simulations. The results show that our model produces intuitive results and agrees with the Bayesian probability calculus.