The improvement of a hypergraph-based model against information leakage by inference