"5G Networks Vulnerable to Adversarial ML Attacks"

A recently published paper called into question the security safeguards built for 5G networks. According to a team of academic researchers from the University of Liechtenstein, a network jamming strategy could allow an attacker with no insider knowledge to disrupt traffic on next-generation networks, even when advanced defenses are in place. The team said the key to the attacks is using an adversarial Machine Learning (ML) technique that does not rely on prior knowledge or reconnaissance of the targeted network. The current approaches to controlling network packets are no longer effective as 5G networks are deployed, and more devices start using those networks to move traffic. The researchers observed that several carriers plan to deploy ML algorithms that can better classify and prioritize traffic to make up for this. These ML models turned out to be the attack's weak spot since tricking them and changing their priorities would allow attackers to change how traffic is handled. The researchers highlighted that a "myopic attack" could bring down a 5G mobile setup by flooding the network with garbage traffic. The fundamental concept behind this attack is to slightly alter the data set. An ML setup would be fed unexpected information by sending a data packet request with extra data. Over time, those malicious requests can change the ML system's behavior to obstruct legal network traffic and ultimately slow down or halt data flow. Although the type of 5G network and ML model used will affect real-world outcomes, the team's academic experiments yielded impressive results. Using the method that required no prior knowledge of the carrier, its infrastructure, or ML technology, the network was brought down in five of the six lab experiments. This article continues to discuss the study on the exposure of 5G network infrastructures to adversarial examples. 

SearchSecurity reports "5G Networks Vulnerable to Adversarial ML Attacks"