Robotics uses sensors, control systems, manipulators, and software to efficiently manage and control devices for specific applications. These components communicate with each other by using protocols.
The danger is that if data in the network communication is not encrypted, it can be easily intercepted by an unauthorized entity to either eavesdrop on the network or inject malicious code into it.
Therefore, cybersecurity for robotics demands higher flexibility beyond current security technologies. Cyber attacks in robotics generally fall under two categories: An endpoint compromise and a network communication-based attack. An endpoint compromise sees a controller unable to control the robot. In contrast, the network communication-based attack encourages an attacker to either eavesdrop or inject malicious code into the network.
Physical access is more to the network-communication-based attack, so they are much more feasible than endpoint compromise attacks. Below, we list common cyber security risks and attacks faced by robots.
1. Intention Modification Attacks
An intention modification attack is performed deliberately to affect the actions of a robot commanded by the controller. This attack alters the message while the packets remain in transition mode. Specifically, packet headers are modified by an adversary to either direct the packets to another destination or to modify the data present on target machines. Denial of Service attacks represents a class of Intention Modification attacks. In such attacks, the robots͛ network interface is overwhelmed with TCP traffic. It may result in either of the following.
a. robot halting: a physical indicator that the robot has been overwhelmed by a Denial of Service attack. It may also lead to erratic movements in the robot. The robot may halt repeatedly and for different durations. Also, the speed may vary.
b. Delay in responding to direction commands: A robot under Denial of Service attack displays a delay during the transition from low to high speed. It may not respond instantly to various navigation commands.
2. Intention Manipulation Attacks
In an intention manipulation attack, the attacker reconstructs the message transmitted from the robots to the controller. These are also known as feedback messages, as they respond to the controller’s input. They may be in the form of video clips or readings. As the controller’s intention is authentic, there is some level of difficulty in performing these attacks. However, detecting or preventing such attacks may be difficult if executed correctly. If the manipulated feedback is believed to be legitimate, it may lead to unfavorable consequences. Most of the robots are governed by communication networks, so they are highly vulnerable to manipulation attacks. A worm may be written to manipulate components in the robot and spread over the network without human intervention.
3. Hijacking attacks
It is an attack wherein the adversary takes control of the communication between two endpoints. If the endpoints were believed to be the controller and the robot, it is possible that the adversary disregards the controller’s intention and executes unethical actions. The hijacking may temporarily or permanently take control of the robot and disrupt the services for a few hours or irreversibly. The hacking of teleported surgical robot Raven II is an example of a hijacking attack on robots.
Two kinds of attackers may carry out the attacks mentioned above. They are:
- Network observer: An adversary to eavesdrop or snoop on the information transmitted between a controller and a robot. He may be involved in collecting information and introducing unreliable information into the communication network while appearing benign to both parties.
- Network Intermediary: An adversary who positions himself between the controller and the robot, thus preventing confidential communication between the ends.
Mitigation strategies
With security breaches knowing no bounds and the domain of robotics being vulnerable to so many insecurities, it is necessary to prevent such attacks before they take a toll on us. Several methods have been put forward for identifying attacks and mitigating the threats to secure a system.
- Communication Robustness: The transmission of commands from a controller to the robot and feedback from the robot to the controller require a transmission medium. It is this transmission medium that is vulnerable to most attacks. Ensuring that a layer of security is provided over the channels for information being transmitted will certainly reduce probable insecurities. Since the communication is dedicated, encryption and introducing authentication mechanisms on transmitted data will restrict modification, manipulation, and hijacking attacks.
- Data Distribution Service in ROS: In Robot Operating Systems (ROS), messages may be transmitted without encryption, encouraging eavesdropping. However, integrating data Distribution Service (DDS) as a transport layer will install plugins that ensure authentication, access control, and cryptography.
- Authentication Mechanism in YARP: When using YARP, the entire infrastructure may be revealed. However, an authentication mechanism may be introduced in the YARP by ensuring key exchange. Port monitoring and arbitration may ensure proper encoding and decoding of transmitted data.
- Securing the Cloud: Cloud robotics is an emanating field that has robotics embedded into the cloud computing environment. It relies on cloud storage and other internet technologies of the cloud infrastructure. It leads to the enhancement of memory, computational power, and interconnectivity for robotics applications. Sensors collect the data and upload the corresponding information to a remote computation center. The information is processed and may be shared with other robots.
- Communication Buses: Communication buses may also provide secure communication. Unlike traditional buses, communication buses are Ethernet-based and can use TCP/UDP/IP features.
