The possibilities of Artificial Intelligence help companies solve or mitigate security problems. This includes both physical security, such as on company premises, as well as technical security, like in corporate networks (cybersecurity). Excerpt from my contribution, which what published in the magazine Security Advisor.
Introduction
How does one teach a computer program to recognize intruders sneaking around on company premises? Can a computer program identify an intruder based on sounds or motion alarm reports that sensors report?
Personally (long-term experience in software development) I would certainly have a hard time writing a program that can solve such tasks based on my specific, pre-programmed requirements.
To answer these questions today, only one question needs to be answered: Can a person do this? The answer is obviously yes. Proof: Security personnel can clearly identify intruders by looking at monitors or patrolling the grounds. What works in real life also applies to networks, such as a company's internal network. There, security specialists determine whether an attack from outside (or occasionally by malicious employees) is to be assumed.
Artificial intelligence learns through examples. Concrete specifications do not have to be programmed in. Rather, examples suffice.
How does man do that? Simplest answer: With his brain. Complete answer: Nobody knows it. Anyway, man learns from examples. In early years of life these examples are especially presented by the parents of the child. This process we call then education.
With the help of modern methods of artificial intelligence, machines can do the same as humans or even better. At least it is certain that a machine does not get tired, but a human already does. Machines also always want to carry out an assigned task, in contrast to humans. Machines are cheaper than employees.
To get an AI system that satisfactorily (or well or very well or excellently) solves a highly complex task, one only has to feed it a lot of examples. Just like with humans. Once an AI system has reached a certain level of intelligence, then for new tasks just a few examples are enough. Just like with humans.
Due to the nature of current AI algorithms I believe they replicate the functioning of the human brain. So far no one has shown me that my theory is wrong.
Security scenarios
To identify intruders, signals must be classified. These signals come for example from cameras, microphones or motion detectors.
A classification is equivalent to file drawer thinking. Thus, people on company premises could be distinguished as follows:
- Innocent passerby or customer
- Janitor or handyman
- Burglars
To recognize a burglar disguised as a craftsman, a specific alarm should also be triggered when a person is recognized as a craftsman. With the help of a Trust Value, it can be controlled whether a quiet or loud alarm is installed, possibly with direct and automatic calling of the police or additional security personnel.
Analytical results that occur over a period of time can be viewed in sections to identify underlying events that are relevant in total. This way, burglars can also be recognized who always dance just a little out of line with behavior that would be attributed to an unobjectionable contemporary.
The exact same procedure that can be used for detecting intruders on company premises can also be used for detecting hackers in IT systems. The difference lies in the sensors, i.e., input data. What is a camera on the company premises is, for example, the Firewall in the computer network.
Wherever sensors and other input data exist, whether physical or software-based, AI algorithms can be used for security purposes. Input data without sensorics occur in contact forms, or when calling a URL that hackers like to add security-critical parameters.
Recommendations
Those who want to enter completely unsensible data into a third-party system like ChatGPT are welcome to do so. Often, however, the data is personally identifiable, or it's even business secrets. Sometimes the data isn't extremely critical, but should still not fall into the hands of outsiders for competitive reasons.
A system like ChatGPT is also very slow and chargeable. Quickly, even with moderately intensive use, several hundred or thousand euros per month in costs only arise for using the OpenAI interface.
Better are local systems, i.e., AI systems, that run on their own computers. Servers or providers located in the USA generally have a problem with data protection (keyword: Privacy Shield – Schrems II). These data protection problems dissolve into thin air with local systems. Even business secrets would probably be given by every reasonable thinking person to their own system, rather than distributing them extensively throughout world history.
High-quality AI systems can already be built with little effort. Feel free to contact me if you need anything.
The image accompanying this post what not only created with the help of a locally running AI on my system, but it what also upscaled to larger image dimensions nearly loss-free using another local AI (which I use among other things when articles are printed on paper). This type of image scaling can also be used for images from surveillance cameras.
My contribution has appeared in Security Advisor and is quoted above in other words.
Key messages
Artificial intelligence can be used to solve security problems, both physical and technical.
AI learns from examples, similar to how humans learn.
AI systems can classify signals from security devices to identify potential threats.
AI algorithms can be used for security purposes by analyzing data from sensors, like cameras or firewalls.
Optimized local AI systems are better than cloud-based ones for data protection and cost-effectiveness.
High-quality AI systems can be built easily and used for various applications, including image processing.
My name is Klaus Meffert. I have a doctorate in computer science and have been working professionally and practically with information technology for over 30 years. I also work as an expert in IT & data protection. I achieve my results by looking at technology and law. This seems absolutely essential to me when it comes to digital data protection. My company, IT Logic GmbH, also offers consulting and development of optimized and secure AI solutions.
