Posted on February 19, 2020 at 6:21 PM
In a share demonstration of another level hacking proficiency, some hackers have just made two Tesla cars accelerate to 50 MPH. The hackers softly altered the speed limit by fooling the Mobileeye EyeQ3 system of the car. They did it in such a way that another nearby driver would not notice.
This latest cybersecurity demonstration from McAfree is proof that adversarial machine learning could pose a potential danger to autonomous driving systems. Those looking to commercialize autonomous driving technology have been alerted to the dangers of doing so because there are lots of loopholes the hackers can capitalize on.
The Mobileye EyeQ3 camera systems are installed on autonomous driving cars to help with reading speed limits. They read signs displaying speed limits and send that information to some brands of Tesla’s autonomous driving cars.
Researchers placed very tiny stickers on the speed limit signs, which were almost unnoticeable to other drivers. As a result, the signs were read as 85 instead of 35, when the sticker was placed on the sign. Both the 2018 Tesla S models and X models sped up to 50 MPH.
How Mobileye camera technology was exploited
It appears that the goal of the security researchers was not to hack into the Tesla S and X models themselves. They were interested in the Mobileye camera and how to confuse its judgment. The researchers succeeded in confusing the camera to read and interpret the wrong speed limit. With this information and the limitation of the camera, it shows that hackers could exploit other areas using real-world scenarios.
The Mobileye web page stated that there are over 40 million cars that currently have the Mobileye camera technology installed.
The technology makes use of a machine vision algorithm to deliver advanced collision avoidance and driver assistance systems. Tesla said the main goal here is to keep the car safe on the road.
Two researchers at McAfree, Shivangee, and Povolny, began a research project to find out how to set up an adversary machine learning technology, which is equally called model hacking. They want to confuse and manipulate the Mobileye camera to send the wrong information to the autonomous driving system.
However, the Tesla vehicles did not completely depend on the information it receives from the cameras. But the researchers expertly figured out that they could mislead the camera and give it a wrong judgment on the road signs.
They decided to use electrical tape to tamper with the numbers on the speed limit, making them look different even to the physical eyes. Unfortunately, about two of the Tesla models fell to the deceit, which forces the cars to accelerate up to 50 miles per second.
Tesla leading the chase for autonomous driving
Tesla already has a reputation for pushing up for the mass production of electric and autonomous cars. The company is at the forefront of this revolution in-car technology. However, time after time, researchers have proven that commercializing these types of cars is likely going to be devastating, considering the faults researchers have found already.
Tesla has been running a series of proofs and tests on its latest models of electric cars and autonomous riding cars. It recently invited hackers to try and hack into the cars’ electric systems. All these are meant to test the readiness of its new car models to meet the needs of the fast-changing auto industry.
There have been some failures in some of the tests carried out so far. For instance, the car windows broke when a steel ball was thrown at them during the Tesla Cybertruck test.
In another failure, researchers discovered that the Tesla S model had some security issues in 2018. But the issue has been patched.
Since then, Tesla has maintained a good record after a series of tests, when it comes to security. Just recently, the car manufacturing giant, in its Pwn2Own initiative, set up a bounty for hackers to evade its Tesla Model S security layer. The company offered a reward of $500,000 to any hackers or hacking group that will be successful to compromise the cars’ security systems.