We are increasingly coming across the term “artificial intelligence” (AI). Whenever we come across this term that has become a fashionable phenomenon in the media, there are regularly reports about self-thinking AI, fully autonomous means of transport or other similar applications. For many of us, it can be hard to figure out what is actually meant by the term. However, the application areas of AI that seem to be particularly popular in the media are in reality only a very small segment of the actual application field of AI. Many a user would probably even be surprised to know in which areas AI is already being implemented today. In this article, we would like to take a closer look at the term and explain what role artificial intelligence can play in the field of IT security.
What is Artificial Intelligence?
The term artificial intelligence can be difficult to define. In general, it refers to technologies that are capable of mimicking or imitating certain human-like behaviours. A distinction is made between strong and weak AI. According to this definition, strong AI would be a system that has some kind of consciousness and whose intelligence equals or even surpasses that of humans. However, these goals are still considered visionary to this day. Weak AI, on the other hand, are those systems that are able to replace cognitive abilities and thus solve very concrete and clearly defined problems. Their goal is not to imitate humans and their consciousness, but merely to simulate intelligent behaviour tailored to the problem at hand. This type of AI is already used in numerous areas of our daily lives.
AI can be subdivided into various sub-areas, each of which builds upon each other. These include machine learning and deep learning, for example. Machine learning enables IT systems to recognise patterns and regularities based on existing data and algorithms. The system can then use these patterns to develop solutions. This means that the IT system generates artificial knowledge from experience. Deep learning builds on machine learning. This learning method is based on the way the human brain works and uses neural networks and large amounts of data. For the AI system, this results in the ability to make its own forecasts and decisions based on the available data.
Fields of Application of Artificial Intelligence
In the media, there are often reports of AI systems in connection with self-driving cars or humanoid robots like Atlas. However, the most common use cases are definitely less spectacular, but no less exciting. Some AI systems we even use on a daily basis. These include, for example, speech recognition software, semantic search engines or knowledge-based systems such as Siri, Cortana etc. AI-controlled chatbots are also becoming increasingly popular in the area of customer support. Artificial intelligences also frequently take on tasks in marketing, the medical field, can be found in the form of non-player characters (NPC) in video games and even create pieces of music and other works of art. By now, AI Systems are finding applications in really any field.
Artificial Intelligence and IT Security
Artificial intelligence has also become an integral part of cyber security, both as a valuable aid and as a potential threat or gateway for hackers. Currently, three major intersections between AI and IT security are the focus of intensive research: attacks by AI, IT security by AI, and IT security of AI.
Artificial intelligence can be used as a useful tool for IT security, and its applications are extremely diverse. For example, they can be used to filter unwanted spam and web content. AI also plays an essential role in system monitoring and the detection of anomalies and fraud patterns. For example, AI systems can be trained to detect activities that deviate from normal user behaviour. In this way, AI can then detect potential vulnerabilities and security holes. AI can also be used to automate certain reactions to malware or viruses. If an infection occurs, this infestation can be detected at an early stage by an appropriately trained AI and measures to limit the damage can be initiated immediately.
Unfortunately, AI systems do not always serve cybersecurity, as they are also increasingly used as attack tools by hackers. For example, AI can be used to systematically detect vulnerabilities. They are also capable of disguising malware and carrying out attacks on other AI. In somewhat more sophisticated attack scenarios, AIs can also easily collect large amounts of data, e.g. personal data, and use this data for extremely subtle phishing, spear phishing and even vishing attacks. At this point, a deep-voice algorithm can use short speech samples to deceptively imitate a person’s voice, for example, in order to fake their identity on the phone.
In the field of IT security for AI, the question is how AI systems can potentially become a gateway for hackers. This involves examining the ways in which an attack on AI systems could take place. For example, an AI system can be manipulated through so-called data poisoning. In this case, attackers deliberately introduce manipulated data in order to contaminate the system’s data records. This can then lead to the AI’s predictions and decisions being falsified. The attack usually takes place over a longer period of time. The AI’s training data is only slightly changed so that the attack remains as unnoticed as possible. To avoid data poisoning, the IT infrastructure around the system should be protected particularly carefully. The uncontrolled learning of an AI system also poses a significant risk from an IT security perspective. The learning process should therefore always be monitored.
The use of AI offers immense opportunities in IT security, but also poses a multitude of risks. They support the analysis of large amounts of data, their evaluation, the creation of diagnoses based on them and their use to defend against cyber attacks. This simplifies countless work processes. In this context, time and cost savings in the manual monitoring of threats are also highly significant. However, careful protection and good monitoring of the AI system’s learning and decision-making processes are crucial here.
More articles on the topic of IT security can be found in this blog under the tag IT security.
Responsible for the content of this article is Stéphanie Bauens.