Deep neural networks (DNNs) have been increasingly adopted in many mobile applications involving security sensitive data and inference models. Therefore, there is an increasing demand for secure DNN execution on mobile devices. Catering to this demand, hardware-based trusted execution environments (TEEs), such as ARM TrustZone, have recently been considered for secure mobile DNN execution. However, it is challenging to run DNN models in TrustZone, due to the stringent resource and performance limitations posed by the mobile TEE. We develop HybridTEE, a novel hardware-based security framework to securely execute DNNs in the resource-constrained local TEE (i.e., ARM TrustZone), by offloading a part of the DNN model to a resource-rich remote TEE (i.e., Intel SGX). HybridTEE strategically divides the DNN model into privacy-aware local and remote partitions by employing two privacy-oriented metrics based on object recognition and Scale Invariant Feature Transform (SIFT). Also, it builds a trustworthy communication channel bridging TrustZone and SGX to enable secure offloading of the DNN model between the two TEEs. Our security and performance evaluations on real hardware systems show that HybridTEE can ensure the security and privacy of the DNN model with superior execution time compared to the non- TEE baseline.