"Deeper Defense Against Cyberattacks"

A KAUST team of researchers has developed a method to improve upon the detection of malicious intrusions to combat the growing threat of cyberattacks against industrial control systems. Internet-based industrial control systems are largely used in the monitoring and operation of factories and critical infrastructure. These systems have relied on costly dedicated networks in the past, but they have been increasingly moved online, thus making them more inexpensive and accessible. However, moving these systems online has also made them more susceptible to being attacked. The researchers explained that conventional security solutions such as firewalls and antivirus software are not appropriate for protecting industrial control systems because they have distinct specifications. The complexity of industrial control systems also makes it difficult for even the best algorithms to detect unusual activity that might indicate an intrusion by malicious actors. Deep learning is a branch of machine learning proven to be adept at recognizing complex patterns. Deep learning is powered by artificial neural networks and is trained rather than programmed. Different examples are provided to the deep learning model from which to learn in order to improve its accuracy as it continues to function. The team trained and tested five different deep learning models using data from the Mississippi State University's Critical Infrastructure Protection Center. These were simulations of different attack types, such as distributed denial-of-service (DDoS) and packet injection, on power systems and gas pipelines. The ability of the deep learning models to detect intrusions was compared to state-of-the-art algorithms. The best algorithms typically had an accuracy rate between 80 and 90 percent, while each deep learning model was between 97 and 99 percent accurate. When the researchers stacked all five deep learning models, the accuracy increased to more than 99 percent. This article continues to discuss the stacked deep learning method demonstrated by the KAUST team that offers an improved way to detect hacking into industrial control systems.

KAUST Discovery reports "Deeper Defense Against Cyberattacks"