Egg white powder adulteration identification technology is of great significance to ensure the quality and safety of egg powder, however, the traditional biomolecular detection methods are complicated and time-consuming, and the adulteration identification model for egg white powder is still mainly a directional identification model, which has a limited detection range and can not effectively cover all the possible adulterants, so it is urgently needed to develop a fast, accurate and generalized method for egg white powder adulteration identification. In this study, we introduced near-infrared spectroscopy detection technology and constructed a LOF non-directional identification model. The model is an unsupervised single classification model, and MSC preprocessing and CARS wavelength screening processing are added to the original model to enhance the model's ability to extract spectral features, reduce noise interference, and lower computational requirements. The experimental results show that the detection rate of the LOF non-directional identification model for adulterated egg white powder can reach 93.6%. Its accuracy, precision, recall, and F1 score reach 93.6%, 95.5%, 93.6%, and 94.5%, respectively. For egg white powder with an adulteration concentration of more than 15%, the total accuracy rate (A_AR) of both test sets reaches 100%, and the average detection time (A_ATS) can be as low as 0.001 1 s. Compared to other non-directional algorithms, this algorithm has higher accuracy and is more generalizable than traditional directional models, making it more suitable for identifying egg white powder adulteration with a wide variety of adulteration types in the market. This study can provide a theoretical basis for the subsequent development of a portable near-infrared spectroscopy detector for detecting egg white powder quality.