Security and privacy are of paramount importance to data-driven edge intelligence services. By offloading computation-intensive model portions to the edge server, split inference can empower low-latency and energy-efficient model inference at resource-constrained devices. The split decision is critical to the communication and computation performances of split inference. However, its impact on security issues has yet to be studied. This article first investigates the security-communicationcomputation tradeoff of split decisions in the inference phase. With the emphasis on security concerns, we summarize the threat models of split inference and illustrate two passive attack mechanisms for recovering the input data and private labels. Case studies are shown to validate the security-communicationcomputation tradeoff and investigate the selection of optimal candidate split points according to the specific device capabilities and service requirements.