Artificial intelligence (AI) is the processing of machines and computer systems to stimulate human intelligence; this can include learning, reasoning and self-correction. There are multiple versions of AI and they are most commonly split into two categories; weak AI, and strong AI. Weak AI is a system that is trained for a specific task, for example a virtual personal assistant such as Apple’s Siri. Strong AI on the other hand, is an AI system with generalised human cognitive abilities so that when it is presented with an unfamiliar task it has the intelligence to find a comprehensive solution.
As one of the most era defining developments of the 21st century, it is no wonder that the technology is now deployed in several different industries. Whether this be AI in healthcare, increasing the speed with which patients are diagnosed; or in education, in which grading is automated and students can be taught at their own speed by a virtual programme. It is evident that AI is becoming increasingly integral to our lives and its impact is likely to be considerable.
In this post we will discuss ground-breaking advancements in AI technology; the effect that it has had on hacking abilities; the cyber risk that AI systems face; and suggestions to combat such risks.
Examples of the Most Ground-breaking Advancements
Driverless Vehicles: AI is working inside vehicles that have deployed driverless technology. While machine learning already exists in many of the technologies with which we interact daily, autonomous vehicles will be the biggest turning point for AI. It is likely to be the first time in which humans rely solely on the technology for an extremely complex task.
Robots as friends- At this stage in AI technology, robots do not have feelings and it is impossible to connect with them on an emotional level. A company in Asia however, has taken a major step forward in the development of a robot companion which can understand and relate to human feelings and emotions.
InferVision: In China, there is a shortage of radiologists to review the 1.4 billion CT scans that come through each year to detect early signs of lung cancer. InferVision trained and taught algorithms have been developed to augment the work of radiologists and allow them to diagnose cancer not only more efficiently, but also with greater accuracy.
What can AI Systems do to Further Enhance Hacking Capabilities?
Over the past years we have witnessed a dramatic spike in cyber-attacks. Ransomware as a Service and readily available phishing technologies have facilitated hackers with no prior knowledge or IT skills to initiate financially crippling attacks.
With the implementation of AI, we can only predict that the scale and impact of these attacks will increase. AI can be taught to automatically detect weaknesses in computer code to expose large amounts of public domain and social network data. Additionally, it can monitor emails and text messages and create personalised phishing e-mails for social engineering attacks. Essentially, it has the potential to become an automated tool, replicating what today’s hackers are already doing but with a higher level of accuracy and efficiency. While it is not easy to trace a hack back to its source (making it such a powerful weapon), it is predicted that AI is already being utilised as a hacking tool. This perhaps explains the dramatic increase in cyber attacks over the past five years; a figure that is only forecast to grow.
What are the Risks to AI Systems Themselves?
Considering societies’ increasing dependency on AI to perform complex tasks, the potential consequences of an attack on the AI systems themselves should not be ignored. It is therefore important that businesses and individuals alike understand the risks faced should a successful attack occur.
The following are just a few of the many examples of how an AI system may be compromised:
- Fooling autonomous vehicles to misinterpret road signs
- Manipulating facial recognition systems for phone unlocking and ATM access
- Bypassing spam filters
- Faking voice commands
- Misclassifying AI based medical predictions
- Social network filtering
As exemplified above, a compromised AI system could lead to detrimental consequences. Cybersecurity must therefore be considered at all stages of the AI development process.
As can be seen from the above, AI is adding a new dimension to the ever-growing cyber threat landscape. It is imperative, at the very least, that businesses keep themselves apprised of these developments and ensure that their risk management strategies keep apace.
However, it is not all bad news. AI is also proving to be a potential ally in the detection and prevention of cyber related risks. Cyber security experts are already adopting AI to fight and test IT security defences, utilising multiple AI computing systems to attack their own networks to automatically identify weaknesses in their systems. This methodology has been adopted by corporate technology giants such as Google and IBM. Even the Pentagon advanced research wing is purportedly using this approach. So, AI can take the fight to the cyber criminals by allowing organisations and sovereign states to accelerate incident detection and response.
It is also interesting to note, although in a slightly different context, that the insurance industry is already adopting machine learning within its own ranks. AI is being used in customer service to both augment and replace human agent interaction, algorithms are being applied to customer driving performance data to help inform the development of products and it is also offering increased efficiency and decreased labour costs in insurance claims processing. So, the insurers are more than aware of the benefits AI brings.
Whether it is due to the advancing technologies that AI offers or as a result of more traditional cyber-attacks, it is important that businesses look to the insurance industry for loss mitigation support and pre and post loss response. As the coverages afforded by cyber policies become increasingly responsive to the latest cyber threats e.g. cyber extortion, cyber terrorism or financial crime, it is important that cyber insurance forms an integral part of the overall risk management strategies that businesses adopt. AI is yet another risk area to which the insurance industry will no doubt need to respond.
If you would like to discuss please get in contact at Cyber@Safeonline.com