Mobile tethering allows devices to share Internet connections, yet the sensitivity and privacy concerns of tethering traffic traces make it difficult to release real-world datasets publicly. Synthetic data emerges as a viable solution for preserving privacy while sharing data. Nonetheless, existing approaches struggle to capture the complex multi-user behavior characteristics in tethering traffic and fail to effectively model the diverse range of traffic patterns that occur in various tethering scenarios. To surmount these challenges, this paper proposes TetheGAN—a novel framework designed for generating mobile tethering traffic, which produces high-quality synthetic traces through a User Snapshot Module and a Traffic Merging Module. In the User Snapshot Module, we generate detailed user traffic profiles by applying embedding techniques to metadata fields, transforming attributes into dense vectors for precise representation. In the Traffic Merging Module, we dynamically merge multi-user traces into a cohesive tethering dataset using timestamp-based merging rules. The experimental results on two typical tethering scenarios show that the synthetic traces generated by TetheGAN significantly outperforms existing approaches in terms of distribution similarity metrics, while also meeting user requirements for downstream tasks in terms of accuracy and ranking order.