Self-driving cars, believed to be on track to become a seven trillion-dollar industry by 2050, are likely to be a popular mode of transportation in the near future, as the technology promises safety and efficiency. Significant advances have been made to achieve these objectives. However, the idea of a self-driving car completing its goals without human intervention in the presence of uncertainties is still posed with challenges. These challenges mostly relate to decision-making under uncertainties that could arise from transient faults or deliberate attempts to disrupt the system in favor of adversarial motives. Traditional schemes of hardening the system against known compromises will not suffice primarily because of the sophistication adversarial attempts have come to achieve and the rate at which new attacks are being devised. The problem demands a solution that is resilient to attacks and that would allow a self-driving vehicle to complete its designated task in the presence of attacks or faults. Several attempts have been made towards achieving resilience in autonomous vehicles. To the best of our knowledge, this is the first work to give a thorough review on the resilient approaches adopted in the context and presents a taxonomy of such approaches. The paper also includes a brief introduction to the architecture of self-driving vehicles and its vulnerabilities in the presented context. This paper also discusses a novel approach using N-version programming for resilience.

Self-driving cars are going to be the main future mode of transportation. However, such systems like, any other cyber-physical system, are vulnerable to attack vectors and uncertainties. As a response, resilience-based approaches are being developed. However, the approaches lack a sound attack model that recognizes the attack vectors and vulnerabilities such a system would have and that does a proper severity analysis of such attacks. Moreover, the existing attack models are too generic. Currently, the domain lacks such specific work pertaining to self-driving cars. Given the technology and architecture of self-driving cars, the field requires a domain-specific attack model. This paper gives a review of the attack models and proposes a domain-specific attack model for self-driving cars. The proposed attack model, severity-based analytical attack model for resilience (SAAMR), provides attack analysis based on existing models. Also, a domain-based severity score for attacks is calculated. Further, the attacks are classified using the decision-tree method and predictions of the type of attacks are given using long short-term memory network.

Software-defined networking (SDN) is becoming a predominant architecture for managing diverse networks. However, recent research has exhibited the susceptibility of SDN architectures to cyberattacks, which increases its security challenges. Many researchers have used machine learning (ML) and deep learning (DL) classifiers to mitigate cyberattacks in SDN architectures. Since SDN datasets could suffer from class imbalance issues, the classification accuracy of predictive classifiers is undermined. Therefore, this research conducts a comparative analysis of the impact of utilizing oversampling and principal component analysis (PCA) techniques on ML and DL classifiers using publicly available SDN datasets. This approach combines mitigating the class imbalance issue and maintaining the effectiveness of the performance when reducing data dimensionality. Initially, the oversampling techniques are used to balance the classes of the SDN datasets. Then, the classification performance of ML and DL classifiers is evaluated and compared to observe the effectiveness of each oversampling technique on each classifier. PCA is applied to the balanced dataset, and the classifier’s performance is evaluated and compared. The results demonstrated that Random Oversampling outperformed the other balancing techniques. Furthermore, the XGBoost and Transformer classifiers were the most sensitive models when using oversampling and PCA algorithms. In addition, macro and weighted averages of evaluation metrics were calculated to show the impact of imbalanced class datasets on each classifier.

Increasing the resilience of traffic control systems is a priority for many important cities worldwide. This is due to the ever-increasing problems leading to different failures in such systems. We are witnessing the intensive introduction of new technologies that automatically manage traffic but are exposed to different kinds of attacks. There are also unpredictable increases in climatic changes and the number of cars in many cities. These factors will surely enhance the failure risks of such systems and consequently increase the damage caused by traffic jams and road accidents. In this paper, we introduce a resilient traffic control system that consists of three levels: sensor control, display, and light control. Each level has three (or more) versions and a dynamic voter. Hence, the introduced system is based on diversity and redundancy (replication), called N-versions. We propose two techniques for the introduced resilient traffic control system. The first technique uses N-versions and dynamic voters to vote between the outcomes in each level. The second technique uses N-versions, dynamic voters, and acceptance testing units. The overhead in the second technique is evidently greater than that of the first technique, but its resilience is better. A fine analytical study is conducted and shows that the first technique requires only three versions to reach the optimal results, bounded by 1/15 probability of having a faulty system. The second technique leads to better results, which can determine small probabilities.

Self-driving cars are expected to become a convenient source of transportation in near future. However, achieving a self-driving car at Level 5 still poses challenges. The challenges could be attributed to maintaining desirable functionality under uncertainty while accomplishing a task. This paper proposes a hybrid N-redundancy-based resilient framework for the perception module that integrates abstract sensor fusion and a probabilistic deep learning model to ensure resilient performance. Experimental results using the CARLA simulator demonstrate that our model consistently outperforms existing approaches in runtime efficiency while maintaining functional integrity under attack. Time-based performance evaluations reveal that our hybrid model achieves faster and more reliable outputs, validating its effectiveness in resilient autonomous perception.

Software-Defined Networking (SDN) enables centralized control and programmability, providing a promising foundation for intelligent traffic management and enhanced security. This paper proposes a simulation-based SDN framework that integrates multiple Machine Learning (ML) models to evaluate the effectiveness of cyberattack detection and to manage packet resilience through the analysis of metrics such as throughput, switch delay, controller response time, and security performance indicators. Experimental results demonstrate that the Decision Tree (DT) and Extreme Gradient Boosting (XGBoost) classifiers achieved higher throughput, whereas the Deep Neural Network (DNN) and Transformer models imposed greater computational overhead. In addition, the DT and Random Forest (RF) models showed an accuracy–responsiveness trade-off, with better throughput and less processing time. Experimental results demonstrate high classification accuracy, with DT, RF, and XGBoost achieving near-perfect detection rates, whereas DNN and Transformer models maintained competitive performance despite certain class-specific limitations. The F1-scores validate the robustness of these classifiers. Complementary SDN simulations assessed packet management efficiency, yielding a consistent packet drop rate of approximately 31–32%, throughput ranging from 7.55 to 9.79 packets/s, and average switch delays around 12 ms across models. The findings confirm that the suggested hybrid approach not only detects cyberattacks effectively but also maintains acceptable SDN performance under diverse traffic conditions.