Künstliche Intelligenz

We are increasingly surrounded by artificial intelligence (AI). Autonomous driving or automatic facial recognition are just two popular examples of how AI is changing our everyday lives. IT security is also on the verge of upheaval, as AI has enormous potential for defending against cyber attacks. The attackers do not rest and use AI to optimise their attacks. Despite all the hype, however, it is often overlooked that users today can already benefit from conventional functionalities that are on a par with AI-supported technologies.

An example of such a conventional technology is the Level of Trust system in NoSpamProxy. This self-optimising system learns users’ communication relationships and uses this information to evaluate emails from unknown senders. These and other functionalities are not reinvented by AI but can be implemented more easily and effectively using AI technologies such as neural networks.

Artificial Intelligence Already Enhances Business Processes

Returning to the influence of AI, however, there is a huge number of application scenarios for AI that are not quite so obvious in everyday life,  for example in production and manufacturing. According to a study, plant utilisation can be improved by 20 percent if maintenance work is carried out with AI-supported foresight.* Further examples are the optimisation of logistics operations through context-conscious robots, automated quality control processes or an improvement of the supply chain through sales forecasts.

Customer service and support can also benefit from AI, for example through chatbots, pre-sorting emails using Deep Learning (DL) and Natural Language Processing (NLP), or automating IT support functions. There is another area in IT where the impact of AI is increasing: IT security and attack defence.

The Growing Threat of Connected Systems

The fact that effective protection against spam and malware is indispensable for businesses today no longer requires any discussion – all too often one hears of large-scale cyber attacks resulting in catastrophic consequences. Phishing emails, virus-infected attachments or malware that cripples computers often result in financial losses that can drive companies into bankruptcy. In addition, there is the inevitable data misuse associated with such attacks. The increasing interconnectedness of IT systems and the evolution of the Internet of Things are further reinforcing the need to secure IT infrastructures, as interconnectedness is accompanied by a higher number of points of attack and an increase in potential vulnerability.

Optimizing Email Security Through Artificial Intelligence

The use of AI offers numerous opportunities to improve the security of IT systems, including email security. The transition from pattern-based virus scanners and email gateways to self-optimising systems that automatically learn from previous attacks has already begun. We are in the midst of a change that will fundamentally change the field of IT security: Attack patterns and virus signatures are changing at an unprecedented speed, and it is becoming apparent that even a constant comparison of virus signatures and refreshing of signature libraries will at some point no longer be sufficient to avert threats. Our experience with the self-learning Level-of-Trust system in NoSpamProxy is an important step in this direction.

AI offers criminals new attack vectors.

Cyber-Angreifer

Dual Use: Attackers Also Speak AI

The enemy doesn’t rest. The so-called dual-use potential reminds us that AI can not only be used to improve protection measures, but can also be used by attackers to increase the accuracy and frequency of cyber attacks. The technological arms race is in full swing.

According to one study, there are three possible effects of AI on the development of cyber threats:

  • Extension of existing attack strategies, e.g. through better personalisation
  • The emergence of new threats, such as attacks targeting specific vulnerabilities
  • Altered character of the attacks, i.e. more effective, more efficient and scalable attacks*

As the threat of AI-assisted cyber attacks grows, so does the need for self-learning systems to analyse ever-increasing amounts of data so that not only can response times be shortened, but attacks can be proactively prevented. Machine Learning (ML), i.e. learning a behavioural pattern based on data, is the basis for the development of powerful countermeasures and reliable defence against attacks.

Artificial Intelligence Is Becoming Indispensable

AI-based systems permeate our daily lives, our production of goods, the provision of services and our own learning habits. In addition, AI is increasingly being integrated into processes in which security and data protection play a central role. AI systems will also become indispensable in the future for defending against cyber attacks and protecting IT infrastructures.

Continuous research in the field of AI is the only way to continue to guarantee reliable protection and complete email security. In addition to the level-of-trust system mentioned above, we have a technology partner in Cyren today that also relies on artificial intelligence and machine learning to detect and eliminate even the most difficult and complex threats. Our own applications and technologies will also be continuously optimised using AI.

The dynamics in the field of IT and email security are high. The sensible use of AI-based and self-learning systems is therefore of great importance for the future development of NoSpamProxy. True to the motto The path to hell is paved with good intentions, we examine possible deployment scenarios to see whether and how AI can help in spam protection. In this way, we ensure that our customers are adequately protected against future dangers.

Free Email Security Check

We put your email security to the test in a simulated cyber attack.

*https://www.mckinsey.com/~/media/McKinsey/Industries/Semiconductors/Our%20Insights/Smartening%20up%20with%20artificial%20intelligence/Smartening-up-with-artificial-intelligence.ashx

*https://www.plattform-lernende-systeme.de/files/Downloads/Publikationen/20190403_Whitepaper_AG3_final.pdf