With the rapid development of modern optical communication networks, the security of optical networks has attracted much attention. In order to ensure the accuracy of information and reduce the frequency of abnormal information in optical networks, a particle swarm optimization based method for extracting abnormal signal data from optical network nodes is proposed. Firstly, design a filter bank to perform frequency band processing on the signals of optical network nodes. On this basis, construct decision statistics and use them as the basis for decision processing of each frequency band to achieve node signal enhancement processing; Secondly, based on the empirical modal algorithm, the node signal is decomposed, and effective IMF components are obtained through screening, and their energy is calculated as the signal feature of the node, providing a basis for the subsequent extraction of abnormal data. Finally, the particle swarm optimization algorithm is used to optimize the weights of the BP neural network, and the node signal features are input into the optimized neural network to extract abnormal data from the optical network node signals. Experimental verification shows that this method has a good effect on enhancing node signals, with high accuracy and stability in extracting IMF component energy and abnormal data of optical network node signals.