基于递归神经网络的非结构化数据加密存储

Encryption storage of unstructured data based on recurrent neural network

为实现非结构化大数据的安全存储并提高其抗攻击能力,提出了基于递归神经网络的非结构化数据加密存储方法。通过分块处理医院非结构化大数据,获得输入、输出张量,构建基于LSTM的递归神经网络,生成医院大数据序列;通过向前反馈和向后反馈获得加密数据,将加密后数据包用加密后源数据包的线性组合代替;再设计基于列不满秩概率存储算法,通过加密存储子节点控制器分配加密存储任务。实验结果表明:该方法具有较强的抗攻击能力,存储效率优势更显著。

In order to realize the secure storage of unstructured big data and improve its anti-attack capability,this paper proposes an encryption storage method of unstructured data based on recurrent neural network.The input and output tensors are obtained by block processing of unstructured big data of hospitals,and then the recurrent neural network based on LSTM is constructed to generate hospital big data sequence,and then encrypted data is obtained by forward feedback and backward feedback.The encrypted data packet is replaced by the linear combination of encrypted source data packet,and then the storage algorithm based on column non-rank probability is designed.The encrypted storage task is assigned by the child node controller of the encrypted storage.Experiment results show that this method has strong anti-attack ability,and the storage efficiency advantage is more significant.