Mathematics
2024, 12, 571
8 of 26
Third, a substantial body of work focuses on the application of advanced ML tech-
niques, such as deep neural network (DNN) and LSTM algorithms, to counteract cyber
threats in EVCS [
22
–
25
]. This research spans the development of effective IDS, the chal-
lenges posed by the integration of EVCS with emerging technologies such as 5G, and the use
of techniques such as WCGAN combined with DL classifiers for enhanced attack detection.
Finally, the realm of privacy preservation in EVCS has been addressed through research
into adaptive, differentially private federated learning mechanisms [
26
]. This is crucial
in optimizing privacy while maintaining data utility in federated learning environments,
presenting solutions to balance privacy and model performance.
In contrast to earlier models that primarily focus on Distributed Denial of Service
(DDoS) attacks using datasets like IoT-23, our model’s ability to classify more complex
and diverse attacks such as injection, scanning, malware, and Man-In-The-Middle (MITM)
sets a new benchmark in the field. Additionally, the utilization of the Edge-IIoTset dataset,
which captures real-world traffic, further validates the practical applicability of our model
in real-time IoT environments, a distinct edge over the CIC-IDS2018 dataset used in some
prior studies (Table
1
).