An assurance case is a body of evidence organized into an argument demonstrating that some claims about a system hold. It is generally developed to support claims in areas such as safety, reliability, maintainability, human factors, security etc. Practically, both argument and evidence are imperfect, resulting in that we can hardly say the claim is one hundred percent true. So when we do decision-making against assurance cases, we need to know how much confidence we hold in the claims. And the quantitative confidence would provide benefits over the qualitative one. In this paper, an approach is proposed to assess the confidence in assurance cases (mainly arguments) quantitatively. First we convert Argument Metamodel based (ARM-based) cases into a set of Toulmin model instances; then we use Hitchcock's evaluative criteria for solo-verb-reasoning to analyze and quantify the Toulmin model instances into Bayesian Belief Network (BBN); running the Bayesian Belief Network, we get quantified confidence from each claim of the assurance case. Finally, we illustrate our approach by using a simplified fragment from safety cases and discuss several future work.