01860nam a22001097a 4500100007500000245015800075260002700233300000900260520140000269650004801669856003301717  aEngr. Iraj Memona2-MCSE-04aSupervisor - Prof.Dr. Fareed Ahmed Jokhio  aAn Effective Distributed Denial of Service (DDOS) Attacks Detection Using Resilient Back-Propagation and Gradient Descent Training Algorithms (ME Thesis)  aNawabshahbQUESTc2024  a60p.  aABSTRACT

In present time where everything is digital and various devices are connected with reliable and fast computer networks. Although networks are secure and can handle various types of networks attacks. However, it is challenging to detect Distributed Denial of Service (DDoS) attacks. Because these attacks are launched from trusted devices. During a DDoS attack, several requests are sent to a server from trusted computers or devices and server may go down completely or it can have delay in services which it is providing. This may cause serrious problems for the users of the system. Timely detection of such type of attacks is essential. This research work explores three Artificial Neural Network (ANN) models: the Feed forward Neural Network (FFNN), Cascade-Forward Neural Network (CCFNN), and Fitnet Neural Network (FNN) for the purpose of DDoS attack detection. For each neural network three training algorithms are used to training. This research, applied Resilient Back- propagation, Gradient Descent, and Conjugate Gradient with Powell-Beale Restarts techniques. The results of the study demonstrated that the Fitnet Neural Network outperformed the other models in terms of accuracy and required a shorter duration to achieve this accuracy across the three training algorithms mentioned.

Keywords: DDoS, DDoS attacks detection, Feed Forward Neural Network, Sparse Autoencoder  aDepartment of Computer Systems Engineering   uhttps://tinyurl.com/3r9d76pa