No longer just science fiction, the Internet of Things and artificial intelligence are important technologies in ensuring safety in a digitally connected city of the future.
From world-beating chess machines to self-driving cars, artificial intelligence (AI) has captured the imagination in recent years. What was once science fiction has begun turning into science fact of late.
At the same time, the Internet of Things (IoT) is abuzz with an increasing number of sensors and data collection points, ranging from the smart devices carried and worn all the time by users to cyber sensors that watch for unusual online traffic in a network.
Together, AI and IoT have the potential bring about a dramatic transformation in the quality of life for citizens in a smart city. By merging the physical with the cyber, the connectivity among real world objects and people in an urban environment is closer than ever.
The knowledge arising from the Big Data generated each day – from the health statistics collected by a smartphone user to the stress levels on urban infrastructure such as subway train tracks, for example – is unprecedented.
The amount of digital data generated in a year is growing from 4.4ZB in 2013 to an estimated 44ZB in 2020, according to research firm IDC1. That represents a growth of 40 per cent a year. With that data, many smart city solutions can be developed. From ones that find the best routes for garbage trucks in a neighbourhood to those helping city planners determine where to locate factories away from residential areas, fresh insights can be derived from the data are growing.
This is helped by increasingly sophisticated AI. With machine learning, it can find new patterns and logic, zeroing in on an answer to a tough question, say, the optimal way to transport the largest amount of people during peak hour with the smallest carbon footprint.
Indeed, by melding the physical and cyber worlds, city leaders are increasingly able to plan and optimise the way their city is run. In a digitally connected society, the benefits to citizens are clear.
At the same time, as opportunities flourish for citizens, so will openings be created for cyber criminals and state-sponsored threat actors. As more facets of a population are available digitally, the risks faced by citizens are higher as well. While remote access to the physical world may bring convenience, it also enables cyber attackers to gain easier access to critical systems.
In 2010, the world caught a first glimpse of Stuxnet, a malicious programme customised to attack the industrial systems used to operate Iran’s nuclear programme. It looked for certain control systems in a nuclear plant, then changed the settings to sabotage them.Believed to be written by state-backed authors, the software has been taken apart and reworked by others so it can now cause more damage elsewhere on machines that are not protected.
In the years since, the sophistication of malware targeting industrial systems, often those controlling critical infrastructure such as power plants or telecom networks, has only increased.
At the same time, the setting up of IoT and digitalisation of business processes have meant more connected systems over the past three years. While many also have more security in place today, the digitalisation and “sensoring-up” of systems may provide an attacker with a way to create major damage should he be able to connect to these systems remotely. This could be carried out through an exploit from other connected systems, for example.
Alternatively, an attacker could choose a softer target to create mayhem in a city. In a rush to deploy sensors and connected devices in the city, many planners may not have had set up an adequate level of security in place to prevent them from being hijacked.
One of the biggest distributed denial of service (DDoS) attacks in late 2016 occurred with the help of thousands of CCTV cameras and digital video recorders taken over by hackers. These devices were targeted by a malware called Mirai. It looked for devices that had factory- default usernames and passwords, which were then used to send massive amounts of traffic towards an online target to flood it and knock it out of action.
The attack in October 2016 caused massive outage on the Internet, preventing many users from reaching sites and services hosted on popular cloud platforms, such as Twitter, Netflix, Spotify and Amazon.
For many security experts, the attacks are a wakeup call that is long overdue. As cities and homes become more digital – from using smart kettles at home to setting up flood detection sensors in storm drains – answering the need for stronger security has to be a top priority.
Unfortunately, there are already millions of cameras and other connected devices already in the open. Many will still have their passwords unchanged, while others cannot be patched up to prevent future attacks because their manufacturers may not offer the updates.
The potential for a DDoS attack, possibly against a city’s critical infrastructure, has to be taken into account as more IoT devices come online. In an interconnected world with no borders to prevent such cyber-attacks, it is imperative that governments work closely together to tackle such threats.
One of the biggest challenge today for IoT has to do with securing the devices. Increasingly, city planners connecting up their environments have to consider sensors that have security built in. They have to be able to deliver the data they collect securely and be resistant to being hijacked for a DDoS attack on other parts of a city.
As the cost of IoT sensors fall, as with the maturing of technologies, the emphasis of many vendors is on “baking in” security in the hardware itself. This will make each connected Thing on the network a lot less susceptible to be taken over for malicious purposes.
Keeping IoT devices secure is a top priority because these devices can also help detect unusual phenomena in the physical world. The warning they provide can make a difference during an emergency.
For example, a sensor that detects air quality may be able to pick up an unusual particle in the atmosphere that may signal a chemical agent being released in a terrorist attack.
Individual sensors worn by users may also come in use, for example, in sensing a major event. In 2014, during an earthquake in California, wearers of fitness trackers were awoken in their sleep in the middle of night. This caused a spike in the percentage of people awake suddenly at the same time, as recorded on their personal devices.2
The key in future could be finding a way to harness the continuous stream of data users might want to provide to enhance the security of the city they live in. Drivers in Singapore can now volunteer to share videos they record in their car cameras with the police3 to collect evidence for a case.
Smartphones in the hands of members of the public are also important in the event of an emergency. A police app lets citizens easily send videos and pictures to the authorities, say, if they see a suspicious person in public. The increasingly powerful sensors in personal devices make this a new avenue that the authorities can seek greater citizen involvement in.
For this to work, all the data that is coming in has to make sense to the human operator or city leader. During a crisis, he has to be presented with a clear picture of the most important things happening, while having the noise reduced.
This is where AI will play a growing role in the future of safer cities. Facing attacks that may not come from traditional sources or follow familiar patterns, the deep learning that AI is capable of now can quickly assist in deciphering a situation.
Already, AI-driven image analysis now enable law enforcement and homeland security forces to track persons of interest across different locations that utilise different camera systems. Coupled with other command and control systems, the autonomous system can more easily identify and monitor a suspect over vast stretches of physical space – every time he steps into the eyes of a camera.
In the digital realm, AI is making an impact in cyber defence as well. In fighting off future threats, the priority today is not simply looking for known threats but also what is unknown. Zero-day vulnerabilities, the newfound loopholes that expose systems to attackers, are often the weapon of choice of many sophisticated threat actors today.
What is required to counter threats that one may not have faced before is to look out for tell-tale signs of a possible penetration or impending attack. With AI, a defence system filters out the “safe” or normal signs of everyday activity and focuses on actions or behaviour on a network that may point to an infiltration.
The system also unburdens the human operator from false positives. If a sensitive cyber defence mechanism triggers too many false alarms, it may end up being ignored when a real threat occurs. With AI, the system learns what appears to be regular activity and flags actions that may lead to a security breach, for example, someone copying large files from a server onto a USB drive, then deleting the files on the PC.
In an inter-connected city, AI is the smart defender needed to pre-empt attacks, guarantee the certainty of information and carry out secure hardware development. And as the arms race with cyber attackers continue in the coming years, smart cities also have to evolve their defences by tapping on more sophisticated AI.
AI, after all, has been through three big leaps over the past six decades. In the 1950s, the early version was more akin to solving a “toy problem”, where humans had to tell the machine what exactly to do. In the 1980s, as computers became more complex, users could put together a set of rules and let it follow them to find a solution.
The current iteration, in the form of machine learning, is one where AI is able to derive rules from sets of data given to it. So, by repeatedly showing a machine, say, thousands of images of a cat, it will find the patterns to the object and deduce that there may be, say, a 90 per cent chance an image of a cat is indeed one.
However, there is still a limit to the understanding that a machine can have. Researchers have, for example, added noisy pixels to adjust just 1 per cent of an image and a computer would wrongly identify an object, even though the image still looks pretty much the same – of a cat – to a human.
The next big step is for a machine to not just recognise patterns but understand and explain that a cat is identified by its whiskers, face shape, colour and other features. In other words, it works more like a human mind. It will not just be able to identify a pattern but also understand the reason why it can identify it, then repeat that learning to find new patterns.
This means it can set its own rules for cognition. It may be able to identify an object faster, for example, like a human child who can recognise a cat without having to be shown 10,000 images of it. The capacity for deep learning may characterise the next big leap for AI, something that researchers are still striving to achieve today.
For smart cities, the implications are tremendous. The rapid developments of AI, coupled with the learnings for both man and machine in the field, mean that they will shape the lives of citizens in future in more ways than many recognise.
For city planners, taking AI onboard early today means reaping the benefits early. From making sense of all the data coming in from a network of sensors and devices to developing stronger defences against complex threats of the future, it is a crucial technology in a safer city. Visit