6th March, 2020
Whilst the Cloud will continue to remain a cornerstone of virtual computing power, with Fog computing looking to ease some of the burden, Edge computing & technology continues to be heavily researched and invested. So what is driving the growth of Edge technology and where is it heading?
Scalability, excessive power consumption, connectivity and latency are some of the key driving factors of today from a resource perspective but large organisations such as Apple, Amazon and Tesla are realising the importance of how Edge computing benefits the customer not only today, but how it will in the future.
Real-time customer engagement is something we are all too familiar with, although we are unlikely to consider due to how seamlessly integrated they have now become within our daily activities. Consider the facial recognition software embedded within our smartphones. The complexity of the algorithms that are being computed in milliseconds locally within the device hardware is staggering, enabling us to unlock and use our smartphones in the blink of an eye.
Now consider that the smartphone could not do this due to lack of processing power and instead had to relay data to the Cloud, process it and wait for the reply. Although this may not take more than a second or two on a good connection, what would happen if you were in an area without signal? This would not be acceptable and emphasises the drive towards Edge computing.
However, the algorithm to unlock your phone through facial recognition is a rather small computing task in comparison to concepts of the future, or even concepts that are being experimented and trialled today such as Autonomous Vehicles (AVs).
AVs are commonly referred to in levels, ie Level 1 AVs which includes “autonomous” capabilities such as adaptive cruise control to Level 5 AVs, where the vehicle will completely drive itself from point A to point B, allowing the passengers to read, eat or even sleep.
As technology stands today the majority of AV’s on the road are Level 2 or 3, capable of increasing/ decreasing its own speed, adjusting the steering if the vehicle strays from the lane and safely completing an emergency stop. BMW and Volvo have reported that they will have Level 4 or 5 models on the road by 2021.
Whist this may sound like an exaggeration for marketing purposes, the reality is that the technology is almost there, it is the processing power which is lacking. With current estimations predicting that the average AV will have 400-500 sensors and generate an average of approximately 40 terabytes of data for every 8 hours of driving, the processing power required to compute this data in milliseconds is huge.
Furthermore, this information cannot wait for a delay in reaching back to the Cloud for the information to be computed. It needs to react as quickly as a human would, with zero delay in order to be effective. Edge delivers this through extremely sophisticated computing power contained within the AV itself to mitigate latency.
Edge will compute the complicated algorithms being generated by the AV such as the speed of the car, the condition of the road, the weather conditions, the wear of the tyre, the weight of the vehicle and so on, trillions of variables in milliseconds and the react accordingly ie bring the car to a stop and prevent an accident.
A single AV working alone, however, is not enough. There will multiple AVs on the road, with traffic control systems conducting and diverting traffic, all generating enormous amount of data. This data will need to be seamlessly integrated throughout all the other AVs whilst the Edge computes your AVs data in addition to all of the external data being generated, to ensure that each runs harmoniously.
Artificial Intelligence also plays a part in the future of AVs, however, as all of the processing power in the world cannot calculate a deer or ball coming into the road in front of the AV. Edge AI is the integration of the two technologies of the future so that not only can the AVs process large amounts of data instantaneously, but can also identify sudden changes and adapt accordingly.
With the imminent arrival of 5G, the communication between Edge AI devices is ready to be interconnected, minimising latency between communications and allowing for huge computational processing power to be delivered to wherever it is needed.
However, as with all advances in technology, the need for security is also ramping up. As with all interconnected things, Edge AI could potentially be exploited by malicious actors increasing their capabilities and reach for processing large attacks, or by generating misinformation within AVs and causing congestion throughout cities.
The attack may not be so obvious, however, or indeed direct.
McAfee reported a 4000% increase in instances of crypto-mining in 2018, with criminals leveraging the sheer number of unsecured IoT devices in volume as opposed to fewer devices with a faster CPU speed. It is also important to note that crypto-mining is currently considered to still be in its infancy.
It is highly likely that cyber criminals would seek to explore the capability of Edge AI and the opportunity that it could bring to the crypto-mining campaign. The introduction of crypto-mining malware to Edge AI within AVs could potentially increase the CPU usage, resulting in the slowing down of other processes or ultimately brick the CPU.
An increase in CPU usage and a reduction in processing capability would likely cause an increase in decision making for the AV resulting in delayed reactions to scenarios or feedback from other Edge AI devices which could have devastating effects.
Furthermore, consider if the crypto-mining malware was running so high that it bricked the CPU of an emergency vehicle, denying ambulances, fire engines or police from attending the scene of an incident. This would not only undermine the use of AVs but also the emergency services themselves, which could potentially be another threat vector that nation states could seek to exploit to undermine governments and increase instability.
The potential implications of vulnerabilities within Edge AI is a far reaching and daunting concept which will change the threat landscape considerably. Not only is it another threat vector for cyber criminals to exploit, but is would also be another weapon in the arsenal for nation states to deploy in the ever-changing landscape of future warfare.
Therefore, as IoT becomes more ingrained in society and Edge devices start to assist more and more in our daily activities, it is essential that this concept is understood and addressed within cyber security strategies at all levels. Devices should be tested under the most scrutinised conditions and security policies should be adapted appropriately, whilst cyber security solutions should seek to incorporate all Edge devices to provide a comprehensive understanding of the network, minimising dark space and increasing visibility throughout.