Cloud-Dew Computing Cyber Attack Detection using Asynchronous Training for Distributed LSTM-AE

The development of effective and trustworthy detection systems is essential given the serious threat that cyberattacks pose to connected and autonomous applications. A promising strategy to raise these systems' performance is distributed processing using asynchronous data parallelism. It is essential in this situation for addressing the issues of scalability, data privacy, and data heterogeneity that can be implemented with the help of cloud computing, which offers a flexible and economical infrastructure. This article provides a thorough set of state-of-the-art metrics for processing time in cloud-dew computing using asynchronous training in cyber-attacks detection systems. In light of our research on their scalability and reliability, the results show that distributed asynchronous processing in cloud-dew computing can reduce training time and optimize speedups for our proposed Deep Learning (DL) LSTM-AE model when adding more dew devices which are, in our case, Connected and Autonomous Cars (CAVs). The suggested parallelization method is crucial for protecting CAVs from cyber threats and assuring their dependable and safe operation in terms of automated path planning.