SECURING 5G NETWORKS WITH FEDERATED LEARNING AND GAN

The threat landscape of the 5G network is quite vast due to the complexity of its architecture and its use of virtualized network functions. This landscape can be divided into two categories: Attacks against the Access point and Attacks against the Core. This thesis has been dedicated to analyzing the threats that plague the 5G network with a special focus on the access point. The architecture for the access point was simulated with a federated learning environment to not only secure the privacy of the user data but to also present a realistic scenario from which to perceive the 5G network. The main objective of the thesis was to secure the access point of the 5G network in this federated learning environment. This was accomplished by placing an Intrusion Detection System at the endpoint which would classify the data as either benign or malicious. The effectiveness of this model was checked by simulating a malicious user and con- ducting certain adversarial attacks to determine if the model could defend against them. The study was conducted by performing two specific attacks i.e Label-Flipping attack and Genera- tive Adversarial Networks. The attacks were successful and revealed that a new system should be designed and developed that could be resilient against these types of attacks.