Will my EV secure a high rating of safety in Cyber-Physical Systems (CPS) by adopting machine learning?
Machine learning for Electric Vehicle (EV) in CPS:
Electric vehicles (EVs) are quickly gaining favour as a greener and more sustainable substitute for conventional combustion engine automobiles. With the growing popularity of EVs, effective management and control systems are required to maximize their functionality, energy usage, and charging infrastructure. In the context of Cyber-Physical Systems (CPS), here is where machine learning plays a crucial role in boosting the capabilities of Electric Vehicles. Data being gathered from the physical world, analyzed using computational algorithms, and then used to take action is what is meant by the term “cyber-physical systems,” which refers to the integration of computing and physical elements. In the context of CPS, machine learning techniques enable EVs to learn from and react to real-time data, allowing them to make wise judgements and optimise their operations as shown in figure 1.
Range anxiety, or the worry of running out of battery power before arriving at the destination, is one of the major obstacles to the proper operation of electric vehicles. This problem can be solved with the aid of machine learning techniques, which forecast an EV’s remaining range based on multiple factors like battery charge, driving habits, weather, and traffic data. Machine learning algorithms can anticipate range accurately by analyzing previous data and continuously learning from real-time inputs, which lessens range anxiety and enables drivers to schedule their travels more efficiently.
The creation of sophisticated energy management systems is another area where machine learning helps to increase EVI performance. These systems use machine learning algorithms to optimise the use of energy resources in EVIs, taking into account things like battery capacity, power consumption, and the accessibility of charging infrastructure. Machine learning algorithms can dynamically alter an EV’s energy usage to maximize range while assuring effective use of resources by learning from driving habits, traffic patterns, and user preferences.
Additionally, machine learning is essential for the creation of intelligent electric vehicle charging infrastructure. Machine learning algorithms are able to forecast and improve charging patterns by examining data from charging stations, grid conditions, and customer preferences. This involves figuring out the best time to charge an EV based on the grid’s demand, price fluctuations, and the availability of renewable energy. Based on the user’s driving habits and travel routes, machine learning algorithms can also forecast charging station availability and make recommendations for charging stations.
Additionally, EVI can interact and work together in a Cyber-Physical System with other vehicles and infrastructure components thanks to machine learning techniques. EVs can learn from the actions and mistakes of other vehicles and modify their behaviour by using machine learning algorithms. For instance, if an EV meets a roadblock or traffic jam, it can tell other adjacent vehicles of this situation, allowing them to select an alternate route and reduce both their travel time and energy consumption.
Cyber Risk for Electric Vehicle Infrastructure (EVI) in CPS:
Due to the presence of large sensors, communication devices in EVI and in the electrical vehicle increases the system vulnerability. Engineers and researchers have shown significant interest in exploring the different aspects regarding the cyber-physical interaction in terms of cyber safety. Most of the companies are working on a vehicle-to-grid concept to handle imbalance during peak load demand and peak energy generation. After integration of grid and vehicle, the second important aspect is the security of data communication during vehicle-to-grid interaction [8]. During cyber intrusion the attackers can target attack from the grid side or from the vehicle side which may lead to the following three attack scenario:
· False data intrusion
· Intrusion to gain access of vehicle monitoring and control system
· Intrusion to gain access in grid side systems
Much research literature addresses the problem specifying cybersecurity of smart grid but very few highlight security concern related to electric vehicle and grid integration. The EV customers need a faster and reliable security system to save their time, critical information, and personal safety. So basically the interaction of electric vehicle and grid needs an Internet of Things platform which easily gets affected by Mirai malware. The malware hits the IoT systems and uses to form a bot network which is used for Distributed Denial-of-Service attacks. Charging an electrical needs a secure payment getaway to enhance customer privacy and security. Last decade data shows that the many successful attacks were carried out on the payment gateway making customers bears the financial loss. So the EVI must have a private secure payment gateway with enhanced features like a one-time vehicle password or owner’s driving license number to complete the financial transactions.
The most critical sensors which are given first priority to save from external and internal intrusion are shown in red colour in figure 1. The blue colour box indicates the list of sensors that comes second on the priority list and the rest are at the third number of the priority list.
Similarly, the charging station consists of many sensors to interact with the vehicle and grid for economically feasible prices as shown in figure 3. The most critical points related to charging stations are communication link, grid interaction, customer data, and user interface.
So from the above discussions, three intrusion category is identified which can affect the operation of EVI are discussed below.
A. Intrusion from grid side
· Using grid interaction stations: This is the first weak point in EVI where the cyber intruders may try to get access to the EVI or electric vehicle from the grid side. Intrusion like stealth attack, false data injection, and many more can be used by the attackers to gain access of the charging station.
· Using communication network: The easiest way to penetrate any cyber network is to use a communication link. All possible attacks can be performed if the communication network is compromised.
· Using user interface like Supervisory Control And Data Acquisition (SCADA): This is the most common link used by attackers in both grid-side intrusion and vehicle side intrusion.
· Using fake operator profile: Making a duplicate platform similar to the grid operators so that the credentials of charging station operators can be obtained. Using those credentials, the intrusion will become easier for the intruder.
B. Intrusion from vehicle side
· Physical ports available in electric vehicle: Nowadays every modern vehicle is equipped with Universal Serial BUS (USB) Port for data transfer in an electric vehicle. Usually, it is used to obtain regular data regarding battery status, traffic status, efficiency, and other important parameters. Using USB the intruders can target both the customer and the power grid. The USB attack is categorized into four different attacks which are software on the USB device, electrical attack, Reprogrammable Microcontroller USB Attacks, and reprogrammed USB peripherals. In all the four attack conditions the intruders get access to the main control of the electrical vehicle and can find a possible path to enter the grid server using the charging station user interface unit.
· Through wireless communication: To interact with another electric vehicle, monitoring station, and emergency services every electric vehicle needs wireless communication. The most common attack in this category is the Denial of Service (DoS) attack. Using DoS the cyber attackers gain unauthorized access results in unreliable operation of the system. After DoS spoofing is the second most common attack used by intruders to gain unauthorized access to the system through phishing mail and message.
· Through user interface at charging station: For charging vehicles most of the time the customer will choose a charging station. The attackers may use the human-machine interface to get access and control of the vehicle and can use it to propagate the attack to the grid during the next charging/discharging session.
C. Intrusion against autonomous driving
· Through traffic controller station: To create mass destruction, the intruders can gain access to the traffic control station and then by getting access to traffic control of different areas which may lead to an accident.
· Hacking radio stations: For creating traffic jams and havoc in the city by providing wrong information through the hacked radio station. It may affect critical services like paramedical, firefighting, and police.
Role of ML in Cyber security of EV:
The cybersecurity issues posed by these cutting-edge technologies must be addressed as the automobile sector transitions to the era of electric cars (EVs). The danger of cyber threats and attacks grows significantly as connectivity and data interchange between EVs and external networks increase. By offering intelligent solutions for threat identification, anomaly detection, and risk mitigation, machine learning techniques play a crucial role in improving the cybersecurity of electric vehicles.
In EV cybersecurity, threat detection is one of machine learning’s most important functions. Large datasets including recognized patterns of malware, network vulnerabilities, and cyberattacks can be used to train machine learning algorithms. These algorithms can find possible dangers and unusual behavior's by examining real-time data gathered from EVs, charging infrastructure, and connected networks. For instance, machine learning can spot trends that indicate unauthorized access attempts, malicious software injections, or atypical network traffic. Machine learning algorithms can identify threats in real-time and inform EV owners, manufacturers, or security teams to take the necessary action by continuously learning and adapting to new attack vectors.
Another vital component of EV cybersecurity that relies heavily on machine learning is anomaly detection. By examining user profiles and historical data, machine learning algorithms can create baseline models of typical behavior's. Then, these models can be used to spot departures from the accepted standards. Machine learning algorithms, for instance, may immediately recognize these anomalies and raise alerts if, for example, an EV’s network traffic unexpectedly surges or if an unauthorized person gains access to the vehicle’s system. Machine learning algorithms can enhance the ability to detect anomalies and spot potential dangers that can escape the notice of conventional rule-based systems by continuously observing and analyzing data patterns.
Additionally, machine learning strategies can improve the performance of intrusion detection systems (IDS) in EVs. In order to recognize and stop unauthorized access and assaults, IDS is essential. In order to find patterns linked to well-known attack vectors, machine learning algorithms can be trained to examine network traffic, system logs, and sensor data. Anomaly detection and behavior's analysis are two ways that machine learning algorithms might use to recognize complicated attacks that may include several phases or strategies. This proactive technique aids in risk reduction and shields the systems of EVs and the passengers’ security from potential harm.
Conclusion:
In summary, machine learning significantly contributes to improving the cybersecurity of electric vehicles. It allows for real-time threat monitoring, anomaly identification, risk reduction, and intrusion detection. The safety, privacy, and integrity of electric vehicles and their users may be guaranteed by EV manufacturers and security teams by utilizing machine learning algorithms to proactively identify and mitigate cyber threats. The cybersecurity of electric vehicles can be strengthened with ongoing developments in machine learning and cooperative efforts within the automotive sector, encouraging the mass use of this environmentally friendly transportation technology.
References:
1. R. R. Kumar and K. Alok, “Adoption of electric vehicle: A literature review and prospects for sustainability,” Journal of Cleaner Production, vol. 253, p. 119911, 2020.
2. U. H. Ramadhani, M. Shepero, J. Munkhammar, J. Widén, and N. Etherden, “Review of probabilistic load flow approaches for power distribution systems with photovoltaic generation and electric vehicle charging,” International Journal of Electrical Power & Energy Systems, vol. 120, p. 106003, 2020.
3. T. Chen, X.-P. Zhang, J. Wang, J. Li, C. Wu, M. Hu, et al., “A review on electric vehicle charging infrastructure development in the uk,” Journal of Modern Power Systems and Clean Energy, vol. 8, pp. 193–205, 2020.
4. F. Teng, Z. Ding, Z. Hu, and P. Sarikprueck, “Technical Review on Advanced Approaches for Electric Vehicle Charging Demand Management, Part I: Applications in Electric Power Market and Renewable Energy Integration,” IEEE Transactions on Industry Applications, vol. 56, pp. 5684–5694, 2020.
5. S. Acharya, Y. Dvorkin, H. Pandžić, and R. Karri, “Cybersecurity of Smart Electric Vehicle Charging: A Power Grid Perspective,” IEEE Access, vol. 8, pp. 214434–214453, 2020.
6. N. Bhusal, M. Gautam, and M. Benidris, “Cybersecurity of Electric Vehicle Smart Charging Management Systems,” arXiv preprint arXiv:2008.07511, 2020.
7. Hasan, Mohammad Kamrul, AKM Ahasan Habib, Zarina Shukur, Fazil Ibrahim, Shayla Islam, and Md Abdur Razzaque. “Review on cyber-physical and cyber-security system in smart grid: Standards, protocols, constraints, and recommendations.” Journal of Network and Computer Applications 209 (2023): 103540.
8. Z. El-Rewini, K. Sadatsharan, D. F. Selvaraj, S. J. Plathottam, and P. Ranganathan, “Cybersecurity challenges in vehicular communications,” Vehicular Communications, vol. 23, p. 100214, 2020.
9. Girdhar, Mansi, Junho Hong, Yongsik Yoo, and Tai-jin Song. “Machine Learning-Enabled Cyber Attack Prediction and Mitigation for EV Charging Stations.” In 2022 IEEE Power & Energy Society General Meeting (PESGM), pp. 1–5. IEEE, 2022.
10. Berghout, Tarek, Mohamed Benbouzid, and S. M. Muyeen. “Machine learning for cybersecurity in smart grids: A comprehensive review-based study on methods, solutions, and prospects.” International Journal of Critical Infrastructure Protection (2022): 100547.
