Deep Learning in ATM Banking Security
The increase in ATM usage is followed by a surge in fraudulent activities within ATM premises, such as physical tampering, or skimming methods due to inadequate monitoring system in real time. Machine Learning technologies help improve the coverage of ATM surveillance by capturing anomalies in video footages recorded of areas of interest within the facility. However, video footages alone may be insufficient in detecting anomalies, due to factors such as occlusion and non-variation of pose data during tampering vs. normal usage. We have carried out experimental studies in audio anomaly detection to validate whether it can complement anomalies which may be captured by video surveillance. Preliminary results in audio-anomaly detection show an improvement of anomaly recall over video based anomaly detection for certain types of tampering based ATM frauds.
Devendra has started his career in Axis Innovation Lab, the fintech centre of Axis Bank Ltd. where he closely works with the bank’s business teams and startup community from fintech sector to steer cutting edge innovation through collaboration. At Innovation lab, he is currently working with Uncanny Vision, a fintech startup with expertise in deep-learning to deploy computer vision and audio analytics in the field of banking surveillance systems. Devendra has completed his B.Tech & M.Tech in Civil Engineering from Indian Institute of Technology, Kanpur and has been an exchange student at Infrastructure Laboratory, University of Tokyo.