Internet of Things: Encryption for battery-powered devices

What: Security for the Internet of Things (IoT) will need encryption both to secure the communication channels and to enable mutual authentication between the devices and the network (this relies on public key encryption). The problem is that these encryption techniques are computationally intensive and, while for the many devices that are connected to a power source (e.g. in a factory environment) that is not an issue, for battery powered devices in the field the encryption processing will rapidly drain the battery. To overcome this problem MIT researchers have designed a special chip which not only executes the encryption 500 times faster (thereby enabling encryption of higher speed channels) but also consumes 1/400th of the power. The secret of the MIT chip is that they have divided their hardware into special purpose modules to execute specific mathematical functions independently [1].

Implications: While this is still a research project, until chips such as these are incorporated into IoT devices, their security will be vulnerable.

[1] MIT News

Healthcare: Predicting your brain age from your sleep

What: A number of mental health diseases (e.g. Alzheimer’s, schizophrenia, epilepsy, bipolar disorder, major depression) are associated with a brain age that is older than the person’s chronological age. A simple way to determine brain age could potentially serve as a marker for such impairments, but today brain age is measured using expensive MRI processes. Researchers in the US have shown that by measuring the patient’s EEG during sleep they can predict their brain age. Using two large sleep datasets collected in hospitals, they applied machine learning algorithms to determine the difference between brain and chronological age. They validated their model by comparing their analysis for patients for whom they had longitudinal data 5 years apart and their prediction was accurate within half a year [1].

Implications: With an ageing population, cheap predictors of potential impairments become more important. Measuring EEG’s is potentially something that could be done at home with relatively low-cost equipment.

[1] Arxiv

Telecommunications: IoT now accounts for 90% of net subscriber growth

What: According to Chetan Sharma, the well-known mobile telecoms industry analyst, 90% of net new subscriber additions in the US mobile market are coming from connected cars and IoT devices. This is a up from 30% in 2014. While smartphone sales have flattened there has been a sharp decline in tablet connections. Consumers are opting to use Wi-Fi on tablets and connecting them by tethering to their phones when they need a wide area connection [1].

Implications: The long-anticipated growth in IoT is finally starting to occur with connected cars leading the way. Telecommunications companies are scrambling to rollout new network technologies, specifically designed to support IoT.

[1] Fierce Wireless

Artificial Intelligence: Stop your picture being recognised

What: In the wake of Facebook’s privacy issues, people have become concerned about their pictures on the web being identified using facial recognition algorithms. Now, researchers in Toronto have found a way of subtly altering a few pixels in a photograph that will defeat facial recognition algorithms. They trained an adversarial machine learning algorithm on 600 faces which enables it to produces a filter that modifies the face in a photograph in an almost imperceptible manner, but it is enough to defeat the machine vision algorithms because it changes key features. They say that their method will reduce the likelihood of your face being recognised to 0.5%.

Implications: We are entering the counter-intelligence phase of AI: people using AI to defeat AI.

[1] VentureBeat

Transport: The self-driving era is on the verge of beginning

What: Google’s Waymo division has ordered 62,000 Fiat Chrysler minivans to be used for the autonomous taxi service it intends to launch in the US next year. This taxi service will be completely driverless and will initially launch in Arizona, which has favourable weather conditions. On the same day, Softbank announced that it would invest $2.25bn in General Motors’ autonomous driving division which is based on their acquisition of Cruise Automation in 2016 [1].

Implications: For those sceptics who don’t think self-driving cars will take off, the billions of dollars being invested say they are wrong. Google would be spending northwards of $1bn on the minivans which even they would not do if they were not serious.

[1] VentureBeat

Artificial Intelligence: Rapid increase in computing power for AI

What: Since 2012, the amount of computing power used for AI algorithms has doubled every 3.5 months. This has been facilitated by increased parallel processing and the transition from conventional CPU’s to GPU’s to special purpose ‘tensor processing units’[1].

Implications: While OpenAI, the non-profit AI research organisation which performed this analysis, posited that “as long as this trend continues, it’s worth preparing for the implications of systems far outside today’s capabilities”, the results are in fact somewhat misleading.  The OpenAI analysis says nothing about the rate at which energy consumption is growing and ultimately it will be power consumption not computing power which places limits on AI.

[1] OpenAI

Digital Security: Finally, a driver’s license on your phone

What: In 2005, Nokia did a survey and asked people what they never left home without and the answer, not surprisingly, was their wallet, keys and phone. In 2018, this is still the case, because even though you can use your phone for payments, digital keys are relatively rare and you still need your wallet to hold your driver’s license if nothing else. Finally, the US is introducing digital driver’s licenses for your phone. The license is unlocked using the phones biometrics (finger print, face recognition) and the license will allow police to interrogate it via a Bluetooth link without leaving their car. Even if the phone is offline, the authenticity of the license can be validated by special features in the app such as head movement in the photo or a special pattern being displayed when the screen is touched [1].

Implications: In order to maximise their value, standards must be developed for digital driver’s licenses. The International Standards Organisation (ISO) is pursuing this goal so let’s hope that they are quick enough to get ahead of the inevitable rush to implement such licenses across the world.

[1] IEEE Spectrum

Finance: What do bank robbers do when there is no cash?

What: Bank robberies in Sweden have fallen from 110 in 2008 to 2 in 2016. Why? Because cash is rapidly disappearing in Sweden. Having been the first European country to print banknotes (in the 17th century) it is now becoming the first to do without them. Two thirds of Swedes say that they don’t need cash and cash had fallen to only 2% of all payments by 2015. Now less than half of Sweden’s banks actually hold any cash so not only is it difficult for robbers to spend it, but it is also hard to find. Robbers have now turned their attention to cybercrime and stealing high value goods such as Apple products and endangered species. Muggings have also gone up to the point where approximately 1 in 7 Swedes are being mugged per year [1].

Implications: From Sweden’s experience it would be mistaken to assume that a cashless society would have less crime – the value of the transition to cashless is convenience.

Professor Hugh Bradlow is President of the Australian Academy of Technology and Engineering. He is recognised as a global leader in telecommunications technology, including being named by Global Telecom’s Business in the top 100 most powerful executives in the global telecoms Industry two years in a row, and by Smart Company as one of the 12 most influential people in Australian ICT.