Ph.D student Ali Hasan Muosa defended his dissertation entitled “Detecting Internet BGP Routing Anomalies Using LSTM-AE”. A hybrid learning model, the Long Short-Term Memory-based Autoencoders network (LSTM-AE) with dynamic threshold and dynamic features selection through window slides were studied. The proposed model can detect 11 events in all kinds of anomalies through all-kind features (AS-path, volume, and distribution). In fact, it was collected for well-known anomalous internet events over three days at one-minute, five-minutes, and ten-minutes intervals. Thirty-three collectors and ASes could detect anomalies, resulting in 89 datasets; with various timelines and 89900 samples. The defense was accepted with success.
