Sequential recommendation systems aim to recommend the next item that a user may be most interested in based on their historical interaction sequences. However, such sequences often contain inherent noise, which can mislead the system and lead to suboptimal results. Researchers have explored various ways to model items within sequences to reduce the effect of noise. However, detecting noise from the perspective of a single domain alone is often inadequate. Therefore, we propose a novel denoising model, Multi-domain Denoising for Attribute-aware Sequential Recommendation (MDARec), to address the above challenges. First, we design an attribute-aware encoder to enable MDARec to learn richer item information. This information can be used as prior knowledge to guide subsequent denoising. Second, a Multi-Domain Noise Discrimination Layer is proposed, which can discriminate noise information by combining multiple domains to more accurately detect the true location of noise. Notably, MDARec can flexibly and seamlessly integrate with most existing sequential recommendation models to improve their performance. Extensive experiments on four real public datasets show that MDARec outperforms other current state-of-the-art noise reduction methods, and that it can be flexibly applied to various mainstream sequential recommendation models. The source code will be available online at https://github.com/nomanhere233/MDARec .