Shylashree N, Kumar Sachin, Min Hong
Department of Electronics and Communication, RV College of Engineering, Affiliated to VTU, Belagavi, Bengaluru, 560059, India.
Big Data and Machine Learning Lab, South Ural State University, Chelyabinsk, Russia.
Sci Rep. 2025 May 22;15(1):17771. doi: 10.1038/s41598-024-84017-8.
In a linear wireless sensor network (LWSN), sensor nodes are deployed in a linear fashion to monitor and gather data along a linear path or route. Generally, the base station collects the data from the contiguously placed nodes in the same order. When the sensors are deployed closely and with a gradual variation of sensor data along the route, a high degree of correlation exists among the sensed data. The sensed data sequence can be compressed with very low loss in such a situation. In this paper, a joint compression and encryption method for LWSN data is presented. The method is based on the dimension reduction property of an autoencoder at the bottleneck section. The Encoder part of the trained Autoencoder, housed at the Base Station (BS), reduces the number of data samples at the encoded output. Hence, the data gets compressed at the output of the Encoder. The output of the Encoder is encrypted using an asymmetric encryption that provides immunity to the Chosen Plaintext Attack. Thus, both data compression and encryption are achieved together at the BS. Therefore, the procedure at the BS is denoted as joint compression and encryption. The encrypted data is sent to the Cloud Server for secured storage and further distribution to the End User, where it is decrypted and subsequently decompressed by the Decoder part of the trained Autoencoder. The decompressed data sequence is very nearly equal to the original data sequence. The proposed lossy compression has a mean square reconstruction error of less than 0.5 for compression ratios in the range of 5 to 10. The compression time taken is short even though the Autoencoder training process, which occurs once in a while, takes a relatively long time.
在一个线性无线传感器网络（LWSN）中，传感器节点以线性方式部署，以便沿着线性路径或路线监测和收集数据。通常，基站按相同顺序从相邻放置的节点收集数据。当传感器紧密部署且沿路线传感器数据逐渐变化时，感测数据之间存在高度相关性。在这种情况下，感测数据序列可以以非常低的损失进行压缩。本文提出了一种用于LWSN数据的联合压缩和加密方法。该方法基于自动编码器在瓶颈部分的降维特性。位于基站（BS）的经过训练的自动编码器的编码器部分减少了编码输出处的数据样本数量。因此，数据在编码器输出处得到压缩。编码器的输出使用非对称加密进行加密，这种加密对选择明文攻击具有免疫性。这样，数据压缩和加密在基站同时实现。因此，基站的过程被称为联合压缩和加密。加密后的数据被发送到云服务器进行安全存储，并进一步分发给最终用户，在那里它由经过训练的自动编码器的解码器部分进行解密并随后解压缩。解压缩后的数据序列与原始数据序列非常接近。对于5到10范围内的压缩率，所提出的有损压缩的均方重建误差小于0.5。即使偶尔发生的自动编码器训练过程需要相对较长的时间，但所花费的压缩时间很短。
