IEEE Consumer Electronics Magazine - January/February 2023 - 122

Security, Trust and Privacy Solutions for Intelligent Internet of Vehicular Things
Table 1. Dataset description.
S.No
1
2
3
Attack type
DoS attack
Fuzzy attack
Attack free state
No. of messages
656,579
591,990
2,369,868
Table 2. Comparison results.
S. No.
1
2
3
4
5
6
Method
KNN
Random forest
Gradient boosting
AdaBoost
SVM
VGG-16
FPR
5%
5%
4%
5%
4%
0.6%
TPR
93%
93%
93%
94%
95%
96%
Figure 5. Evaluation results of the deep learning
system.
system effectively identifies two different kinds of
attacks in sample messages.
We also show the false positive rate (FPR) of the
deep learning system in Table 2. It can be observed
that the FPR of the deep learning system is 0.6%.
The model results are also compared with other
machine learning algorithms, as demonstrated in
Table 2. It can be observed that the presented deep
learning model achieves excellent results.
From the comparison results of Figure 6, it is
observed that the presented deep learning system
results are good as compared to other algorithms
with a true positive rate (TPR) of 96%.
CONCLUSION AND FUTURE WORK
In this article, a deep learning-based intrusion
detection system is presented for the IoV. The system's
major goal is to provide a solution that
affords STP to communicate entities and secure
vehicle data from malicious attacks. The system
utilized the VGG-16 architecture for the detection
of different cyberattacks, i.e., DoS and fuzzy attacks
in the CAN bus. For training and testing purposes,
we used a CAN-intrusion-dataset, collected at the
OBDport of the vehicles through different external
communication links. The architecture is trained
for the attack patterns to detect malicious attacks.
The experiments reveal that the presented intrusion
detection system significantly reduces FPR.
122
Figure 6. Results of the deep learning system with
other algorithms.
The results are also compared with other conventional
algorithms; among all, the deep learning system
shows high-accuracy results. The overall
accuracy is 96% with an FPR of 0.6%. In the future,
this work might be extended for other deep learning
architectures to enhance in-vehicle intrusion
detection system performance and deliver highquality
solutions for its security.
&REFERENCES
1. P. Bagga, A. K. Das, M.Wazid, J. J. Rodrigues, K.-K. R.
Choo, andY. Park, " On the design of mutual authentication
and key agreement protocol in Internet of vehiclesenabled
intelligent transportation system, " IEEETrans.
Veh. Technol., vol. 70, no. 2, pp. 1736-1751, Feb. 2021,
doi: 10.1109/TVT.2021.3050614.
2. P. Bagga, A. K. Das, M. Wazid, J. J. Rodrigues, and
Y. Park, " Authentication protocols in internet of
vehicles: Taxonomy, analysis, and challenges, " IEEE
Access, vol. 8, pp. 54314-54344, 2020, doi: 10.1109/
ACCESS.2020.2981397.
3. A. K. Sutrala, P. Bagga, A. K. Das, N. Kumar, J. J.
Rodrigues, and P. Lorenz, " On the design of conditional
privacy preserving batch verification-based authentication
scheme for internet of vehicles deployment, " IEEETrans.
Veh. Technol., vol. 69, no. 5, pp. 5535-5548, May 2020,
doi: 10.1109/TVT.2020.2981934.
IEEE Consumer Electronics Magazine

http://dx.doi.org/10.1109/TVT.2021.3050614 http://dx.doi.org/10.1109/ACCESS.2020.2981397 http://dx.doi.org/10.1109/ACCESS.2020.2981397 http://dx.doi.org/10.1109/TVT.2020.2981934

IEEE Consumer Electronics Magazine - January/February 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2023