Artificial Intelligence, or AI, is becoming an ever increasingly useful tool for companies. One way in which companies are employing the technology is through the use of AI to help improve security. This IT transformation within businesses is possible for a number of reasons, including the availability of networks from which an AI program may learn and expand its knowledge. The result is a highly individual security tool that adapts to a company's specific needs and improves itself the longer it is used. Much like human learning makes a police officer better over time due to experience and training, AI security has the capacity to grow and evolve too.
One company at the forefront of this new technology is Amazon, the online retail giant. While it has several AI components for new tasks such as Alexa (its voice-activated digital assistant), Amazon also has an interest in cybersecurity and recently purchased the AI cybersecurity company Harvest.ai. Using algorithms, the AI can identify the most important documents and other data of a business. From there it learns user behavior. It then takes the information about user behavior and what documents and data are important and compiles data loss prevention strategies to guard against cyber attacks. Harvest.ai was founded by former members of several US government security agencies, including the NSA, FBI, and DoD.
The specific product of interest to Amazon was MACIE, which is an AI program designed to monitor a company's network, essentially watching for suspicious user activity and identifying if the user accesses documents for which they are not authorized. Targeting Fortune 100 businesses and other enterprise companies, MACIE was built to work effectively with a cloud-based platform.
Another big name example of AI cybersecurity is IBM's Watson. This project is being trained in security protocols and has entered the beta stage of incorporating clients among several industries. Watson for Cyber Security may now be available to customers for their own company security needs.
Recorded Future is a security firm using machine learning to scan the internet, including black market forums. Its job is to find potential threats, which could include someone trying to sell software exploits or stolen data. An action does not have to be causing harm to be caught and flagged in an AI cybersecurity system. It is like security agents at an event watching someone who appears suspicious, or may be acting like a shoplifter, rather than bringing the person in for questioning before they have done anything wrong.
MIT has designed a project that has the AI learning from a human supervisor. Initially, the AI system could detect 85% of threats, with an anticipated 100% rate over time and further learning. The reason that AI applications work so well as cybersecurity is due to the information they can process. AI programs are able to process a large amount of information in a shorter time than a human can. In fact, over time the amount of data an AI cybersecurity program can search through, record, and analyze would be considered vast compared with a human counterpart. Using a human counterpart to help the AI understand when it gets stuck on the 15% or so of threats increases its usefulness exponentially. The program can then be taught to compare scenarios with specific criteria against actions monitored on a network. Over time, AI can learn and improve on its original cybersecurity, thus adapting to new information, threats, and secure documents.
So how fast can Artificial Intelligence process information and identify threats? According to IBM, Watson takes just 15 minutes to create a security analysis in response to a cyber threat. For a human, this same process would take about a week. That’s only a glimpse at the effectiveness of AI providing a boost to cybersecurity. Big companies have quickly caught on to this impact, and are looking to use it more extensively in the future. Considering the number of threats out there, along with the severe nature of those threats, this is a smart strategy to undertake. Expect far more AI used for security in the years to come—not just for large businesses but smaller ones as well.