Fig 1.
CAN packet syntax.
Fig 2.
Architecture of IDS based on machine learning techniques.
Fig 3.
(a) DBN structure with n hidden layers built with a top-down manner and (b) DNN structure involving the pre-trained wight parameters in n hidden layers built with a bottom-up manner.
Fig 4.
Attack scenario in the connected car.
Fig 5.
Overview of the proposed intrusion detection system.
Fig 6.
The occurrences of a bit-symbol “1” in the DATA field of 8 Bytes, consisting of mode information and value information, at time t.
Fig 7.
Deep neural network structure in the proposed technique.
Fig 8.
Template matching method to find the proper trained parameters.
Fig 9.
Simulation configuration.
Table 1.
CAN packets used in the simulation.
Fig 10.
Intrusion detection performance evaluations with ROC curves.
Fig 11.
Confusion Matrix Results.
Fig 12.
Intrusion detection performances with respect to the number of the layer.
Table 2.
Time complexity in a different number of layers.