Table 1 Related and existing work limitations.

²²

Deep Learning and SOEKS for Intrusion Detection

Requires significant vehicle-specific data for training, which may not always be available

Limited adaptability across different vehicle models without retraining

²³

Hybrid Deep Learning (GRU and LSTM)

High dependency on large datasets for effective performance

Potential overfitting when dataset variety is limited

²⁴

VGG16 Deep Learning Classifier

Reliance on extensive computational resources for training

Performance metrics might degrade with unbalanced datasets

²⁴

Hybrid Deep Learning (CNN and LSTM)

Complexity in model training and hyperparameter tuning

Requires large labeled datasets for optimal performance

²⁶

Analysis of Deep-Learning-Based IDS

Some methods are computationally expensive and require high-end hardware

Limited ability to detect unknown or zero-day threats without retraining