Reposting from Automotive Testing Technology International
By Cesare Garlati
When it comes to testing the components of modern connected cars, of course pen-testing (penetration testing) has its place; however, it is no substitute for solid product development.
In testing, companies often operate under the notion that an identified problem can be fixed or patched. This may be true for some areas of testing, but for security, it is not sufficient. Security needs to be built-in, from the ground up. And that means starting at the hardware layer, which is seldom done today, but which is completely viable given the advancements in silicon and other connected vehicle technologies.
In fact, the prpl Foundation has produced a guide on how to secure critical areas of embedded computing that advocates the use of open, interoperable protocols and APIs, exercising security by separation, through implementing hardware virtualization and anchoring a root of trust in silicon.
Looking back at all of the recent public cases of researchers hacking connected cars, they all share the exploitation of proprietary code. This idea that closed, proprietary systems can work within Internet of Things and connected devices is a myth. In contrast, an open security framework means it has constantly been tested and had many eyes cast over it to ensure its strength.
The second thing they all have in common is that once hackers were able to reverse engineer vendor-specific code to gain access to one area of the system, they proceeded to move around laterally to other networked components. This idea that once an actor can gain access to a non-critical component in a vehicle, such as the entertainment system, and then work their way into a critical area, such as the steering, is scary to think about. But without using the time tested method of security by separation, it is a reality. This separation can be achieved by using hardware virtualization so that although independently they might not be more secure, as a system, one bad apple doesn’t compromise the whole system.
Finally, all of these security controls need to be tied to a root of trust in silicon; this can be a by-product of the hypervisor used in creating hardware virtualization or by some other method. One neat area being explored by prpl at the moment is physical unclonable function (PUF) technology that can extract a unique identifier from the silicon itself, much like a fingerprint, to provide authentication and establish a root of trust.
In summary, pen-testing is important, but it is no replacement for sound product development. Security can only be forged from the ground up in the silicon of connected components themselves. It cannot be added as an afterthought as we have seen time and time again. Testing alone does not make a product secure. From a risk management perspective, testing lowers the risk but doesn’t completely remove it. After all, upon successful testing one can say, “I couldn’t find anything wrong,” which is not the same as saying, “There is nothing wrong.”