The massive streams of Internet of Things (IoT) data require a timely analysis to retain data usefulness. Stream processing systems (SPSs) enable this task, deriving knowledge from the IoT data in real-time. Such real-time analytics benefits many applications but can also be used to violate user privacy, as the IoT data collected from users or their vicinity is inherently sensitive. In this paper, we present our systematic look into privacy issues arising from the intersection of SPSs and IoT, identifying key research challenges towards achieving holistic privacy protection in SPSs and proposing the solutions.