The current power monitoring systems deploy strict security protection strategies on the horizontal and vertical boundaries of each security zone.Despite these protections

the internal network within the local area network remains vulnerable due to minimal traffic restrictions between devices

posing inherent security risks.This study proposes a matrix factorization-based scheme for detecting malicious traffic within the internal network of power monitoring systems.By gathering and analyzing historical network traffic data

a normative traffic model is developed and implemented for real-time monitoring using a whitelist approach.The efficacy of this detection method is validated through experiments

which highlight its high efficiency

proficiency in identifying unknown abnormal traffic

and capability for adaptive model iteration.