In this paper we propose two novel variants of the well-known Steiner tree problem in graphs that are motivated by applications in secure strategic telecommunication network design. Both network optimization models ask for a tree of minimal total edge cost that connects a pre-specified set of terminal nodes to a dedicated root node by optionally including intermediate Steiner nodes. Two types of privacy conflicts between pairs of conflicting terminals are considered: (I) The path from the root to a terminal must not include the conflicting terminal, and (II) conflicting terminals have to be on separate branches of the tree. We develop non-compact integer programming formulations and elaborate branch-and-cut algorithms. We incorporate problem specific valid inequalities that are crucial in order to solve these problems, and establish dominance relationships between these cuts and the induced polyhedra. The effectiveness of the cutting planes with respect to the dual bound and the performance of the exact algorithm are assessed on a diverse set of SteinLib-based test instances.