RESEARCH INTERESTS
Data Mining, Artificial Intelligence (Machine Learning and Deep Learning), Social Computing, Natural Language Processing, Data Science, Internet of Things

Highlighted research work “Anomaly Detection in Dynamic Networks”

Evolution in the nodes along with their attributes leverages the probability of anomalies in the network. We aim to detect anomalies in the dynamic network settings where network structure emerges due to the co-evolution of nodal attributes. In a dynamic attributed network, certain attributes influence nodes and cause them anomalies. In this study, we propose a DEep Co-evolution architecture for anOmaly DetEction (DECODE) in dynamic network settings. Particularly, the proposed architecture models node-attribute embedding learning with the recognized Graph Neural Network (GNN). A Long Short-term Memory (LSTM) autoencoder is trained to reconstruct the learned embeddings. The combinatorial effect of LSTM autoencoders and GNN helps to spot the anomalies by computing network reconstruction errors in terms of both nodes and attributes. We do provide experimentation on real-world datasets that depicts the effectiveness of the proposed architecture.