| Summary: | The wirelessly connected intelligent robot swarms are more vulnerable to be attacked due to their unstable network connection and limited resources, and the consequences of being attacked are more serious than other systems. Therefore, the quantitative assessment of wireless connected intelligent robot swarms network security situation is very important. Factors determining the state of wireless connected intelligent robot swarms network security have characteristics such as mass and diversity, which constantly evolve with time. In fact, network security measurement has multi-level, multi-dimensional, and multi-granularity characteristics. Therefore, properly selecting wireless connected intelligent robot swarms network security measurement parameters and reducing and converging them to quantitative values such that they can enable a true and objective reflection of the network security state is a very challenging problem. However, deep learning is a novel solution to the abovementioned problems; its algorithm gets rid of the dependence on feature engineering and automatically builds a quantitative assessment model of a network security situation with dynamic adjustment as well as self-adaptive and self-learning characteristics. In this study, we propose a quantitative assessment method of wireless connected intelligent robot swarms network security situation based on a convolutional neural network (CNN). Generally, the convolutional layer is used to locally detect and deeply extract features, and the pooling layer is used to rapidly shrink the network scale and highlight the summary features. Using the deep network structure of several hidden layers, the results of quantitative assessment of the network security situation are highly consistent with expert experience. Experimental results show that the quantitative assessment of wireless connected intelligent robot swarms network security situation can be realized by combining the characteristics of a network security index system and CNN. Note that the accuracy rate is 95%, and the calculation results are better than those of other deep learning models.
|
|---|
