The Internet of Things (IoT) is integrating the Internet and smart devices in almost every domain such as home automation, e-healthcare systems, vehicular networks, industrial control and military applications. In these sectors, sensory data, which is collected from multiple sources and managed through intermediate processing by multiple nodes, is used for decision-making processes. Ensuring data integrity and keeping track of data provenance is a core requirement in such a highly dynamic context, since data provenance is an important tool for the assurance of data trustworthiness. Dealing with such requirements is challenging due to the limited computational and energy resources in IoT networks. This requires addressing several challenges such as processing overhead, secure provenance, bandwidth consumption and storage efficiency. In this paper, we propose ZIRCON, a novel zero-watermarking approach to establish end-to-end data trustworthiness in an IoT network. In ZIRCON, provenance information is stored in a tamper-proof centralized network database through watermarks, generated at source node before transmission. We provide an extensive security analysis showing the resilience of our scheme against passive and active attacks. We also compare our scheme with existing works based on performance metrics such as computational time, energy utilization and cost analysis. The results show that ZIRCON is robust against several attacks, lightweight, storage efficient, and better in energy utilization and bandwidth consumption, compared to prior art.