19.08.26 (inclusive)

contractual

Finished product

ICT tasks

Robotics

Neurotechnology and artificial Intelligence

Software/ IS

The current ATM Code SVN video surveillance module provides recording, titration and storage of video stream from ATM cameras, but does not have the function of automatic analysis of what is happening in the frame. As a result, cases of skimming devices being installed, cash trapping being blocked, a customer being held at an ATM by third parties, vandalism, or attempts at unauthorized access to the technical compartment are recognized after the fact — when viewing the recording manually or based on a customer complaint, rather than at the time of the event. Banks need a solution that analyzes the video stream at the device boundary or in the cloud and generates automatic alerts to security services when abnormal behavior is detected. Technical task Computer vision/video analytics technology is required that meets the following conditions: compatibility with the existing ATM Code architecture (platform on .NET 8, RTSP/ONVIF video streams from cameras installed at ATMs and self-service areas); near-real-time operation with edge processing capability on the device side with limited computing resources; scenario detection: installation of extraneous devices on the card reader / keyboard (skimming), physical impact on the ATM case, prolonged presence of a person at the ATM without performing an operation, attempts to access the service area, gathering of several people at the device; transmitting an event via API/webhook to the ATM Code Monitoring module to generate an alert; the possibility of further training the model on the data of a particular bank without transmitting video outside the bank's infrastructure (personal data processing requirement).

Reduction of financial losses of banks from fraudulent actions at ATMs due to early detection of incidents; Reducing the reaction time of security services from after-the-fact review of records to automatic notification at the time of the event; Reducing the burden on staff responsible for manually viewing video archives; Increasing customer confidence in the security of self-service devices.

Губина С.А.