Dr. Tariq Mehmood

Name: Dr. Tariq Mehmood
Designation: Assistant Professor
Affiliation: IBA
Country: Pakistan

 

 

Talk Title: Autoencoders and LSTM for Multi-Class Anomaly Detection
Abstract:Insider Threats and Suspicious Activities are plaguing all important corporate domains. For instance, banking frauds and cyber threats are a major source of concern for both customers and service providers. Many techniques based on machine learning have been used to detect such anomalous or fraudulent activities. In this talk, we will discuss the performance of two deep learning approaches, specifically Autoencoders and Long Short Term Memory (LSTM) for anomaly detection. We focus on multi-class classification of more than 20 classes of anomalies/frauds. We will discuss the results of experiments and draw our conclusions on the positives and caveats of both algorithms.