cars are one of the most exciting and transformative technologies currently under development. Fully autonomous vehicles would revolutionize transportation: decreasing costs, increasing safety, and improving quality of life for hundreds of millions of people who spend hours every day stuck in traffic.
However, autonomous vehicles do present potential security risks which have been recognized for years.
While these warnings have been circulated they have received relatively little attention in the public discourse surrounding autonomous vehicles.
The prevailing attitude seems to be one of “make it work, then make it secure” which is perfectly rational if you assume that fully autonomous self-driving cars will need testing and certification from both government agencies and insurance companies before they are ever sold and consequently do not create any risks prior to the technology being perfected.
Unfortunately this assumption is wrong.
The potential to weaponize autonomous driving technology using existing technologies and vehicles already exists.
Autonomous driving technology is a clear and present danger and it is only a matter of time until a terrorist attack or criminal act using self-driving cars takes place.
Weaponizing Existing Autonomous Driving Technology
To effectively deploy a vehicle as a weapon bad actors do not need fully autonomous driving technology. All they need is technology that allows them to control a vehicle via remote control.
Practical remote control of a vehicle presents numerous challenges: you need to transmit a command signal, you need to control the vehicle based on very limited visual feedback, and you have significant latency between your visual feedback and your ability to direct a change in the vehicle’s operations.
Semi-autonomous flying vehicles (quad-copters, drones) have solved these problems through a combination of machine and human control.
The human controlling the drone sends general commands via remote control, telling the computer to change direction, go up, down, etc, and then the computer implements those general commands through much more precise real time control of the vehicle’s motors while adjusting to feedback from on-board sensors.
This same kind of hybrid remote-autonomous control is fully realizable for automobiles using existing technologies and existing vehicles.
The comma.ai open source self-driving car
) is well suited to developing a hybrid remote-autonomous car by combining its ability to control the throttle, brakes and steering of a vehicle and autonomous assist functionality with a remote control system.
The limited autonomous capabilities of the system would be sufficient keep the vehicle driving stably and responding in real time to road and traffic conditions while the remote operator specified general directional and speed parameters based on live video feeds.
All of this can be done using existing vehicles, existing hardware and existing software. Nothing new needs to be created. Existing technology only needs to be tied together. This lowers the bar in terms of required engineering expertise dramatically.
The only limitation that currently exists is that the comma.ai team has not managed to gain full remote control capabilities at all vehicular speeds yet but they are actively working on the problem and offering cash bounties for solutions.
What Will Happen When an Attack Occurs?
A terrorist attack or a criminal act using autonomous driving technology is a practical certainty. This cannot be prevented because the technology to do it is already in the wild. It is not a matter of if, but when. The only question that remains is what will follow.
If history is any guide the likely outcome of a terror attack using a self-driving car will be media stoked public hysteria followed by a hasty and sweeping government response.
The likely legal and regulatory response is that all new vehicles will be required to have a kill-switch and/or remote control technology that tracks and reports the vehicle’s location at all times, allows the vehicle to be remotely disabled or controlled by law enforcement, and disables or restricts the vehicle’s usage if the connection is dropped for any significant period of time.
Any unauthorized maintenance or modification of vehicles will be made illegal and all vehicles will be fitted with tamper switches that continuously report back to the same central system that tracks vehicle location.
The software that enables autonomous driving will be classified as a weapon, similar to cryptography software in the past, and the development of this software will be restricted to licensed corporations that follow government established security protocols.
Existing vehicles, which can be converted to weapons, will be banned from densely populated areas, effectively requiring the majority of the population to upgrade to new vehicles with enhanced security features.
While the technological potential of self-driving cars is alluring the practical necessities of preventing terrorism and crime will have a profound impact on how this technology is eventually deployed in the real world.
The only way that self-driving cars can be made secure is through new systems of surveillance and central control.
Far from being autonomous these new cars will necessarily be subject to a broad range of new security measures that restrict the individual liberties of their owners.
Autonomous driving technology is fundamentally a weapons technology. Any development that enables autonomous driving is equally applicable to autonomous killing machines.
Ultimately the development and deployment of this technology must be restricted and controlled. The only choices that we have are how those restrictions and controls are implemented and whether or not a catastrophic loss of life must come first.