You may have often seen movie depictions of hackers and been wondering how they get into computer systems, steal stuff and compromise security everywhere. While movie depictions are over-dramatized, the reality often looks more like the typical office job, with hackers using various scripts to find and break into known vulnerabilities of systems and often using the very same people working within an organization as its weakest link, in so-called phishing attacks, that exploit people’s misplaced sense of trust and lack of security awareness to get valid logins, passwords, and so on which are then used to break into more systems from the inside. Often, the people involved remain completely unaware of what is going on, acting unwittingly as accomplices to help hackers gain access, often via seemingly completely legitimate credentials. Security is an area where AI can really help to increase the overall security of modern electronic systems.
I’m going to briefly discuss three different types of security: commercial, home and personal.
In a commercial setting, it is important and obvious to protect your corporate assets. This is because the information and data that a business collects, and all the IT systems that are needed to run the business obviously need to be protected. Security breaches often involve either attempts from the outside — but in many cases, attacks come from the inside.
AI systems nowadays learn how to identify and adapt to possibly malicious attacks that are coming for the outside together with security breaches from the inside. Insider attacks are notoriously more difficult to ward off. By having intelligent monitoring of employee behaviour, insider attacks can be flagged with the minimum compromise to privacy, enabling a good balance of safety against privacy, and by having an AI system doing the monitoring, ensuring impartiality and also reducing the risk of data leaks by people who would otherwise be responsible for such monitoring.
Internal attempts are generally more damaging because they can cause much more economic hardship and problems than external attacks. AI systems can help reduce the risk involved in insider type of attacks by constantly monitoring activities such as failed logins, file copying, emailing, and so on, and assessing the risk involved. Insider attempts may be very subtle and difficult to detect — this is an area where AI can excel over human administrators.
At home we are increasingly seeing the adoption of better security systems that have inbuilt AI, for example, wireless routers that check the connection safety of the router itself and those of your devices. It would be great to know that no one is compromising the security of your future smart fridge or snooping on you via your smart television. AI can also ensure that smart systems like smart CCTV cameras, smart alarms and home automation systems are all integrated into one package and securely adapt to your lifestyle. AI can learn to detect when you are away, know how to turn on different patterns of lights at home when you are away and maintain a place in optimal condition, to deter burglars and make sure that the overall security and safety of its users is increased. Intelligent camera monitoring systems using AI powered video analytics can also detect unauthorised attempts such as theft attempts, or even accidents such as someone accidentally falling down the stairs at home, and may automatically call for help or flag up an incident for human assistance where appropriate. There are even companies that are training AI systems to detect different types of baby cries or dog barks!
Apart from home and office, an increasing AI security application area is personal security. Smartphone security, having an intelligent AI system that will detect malicious attempts and malware that is running on your phone is an increasingly important area. Apart from smartphone security, a lot of companies track and monitor different things about their users, often largely unknown by such users either by design or by being buried under obscure legal language (who actually reads long terms and conditions documents in smartphone apps and does not merely click accept?) and having an AI system that automatically knows what is going on and knows the level of what you want to share and what you want to keep private from these companies will help you automatically cover your tracks.
A personal AI security system will also automatically stop any other AI services from monitoring things that you do not want to be monitored, and in some cases obfuscate and anonymise your data to ensure that there is no hidden algorithmic bias working against you. I think this is going to be a very important aspect in the future as people start valuing their privacy more and more, while AI needs to achieve more trust by clearly explaining the basis of its decision making. Obviously, this needs to be done by striking a balance between the public need with the need for individual privacy.
I think that AI can help us achieve a higher level of security in an ethical way, and with the agreement of how society perceives its trade-off between security and privacy without compromising on the quality of the overall sophisticated safety net afforded by the AI security system.