The main aim is to boost intelligent video analytics software with the extended anomaly analysis option to prevent illegal attacks on ATMs.
AI/ML. Anomaly Detection
The main aim is to boost intelligent video analytics software with the extended anomaly analysis option to prevent illegal attacks on ATMs.
Business Challenge
There is high competition in the VMS solution providers market. Every company tries to offer its current and potential customers video monitoring software with the most attractive functionality sets. Among the most requested features is embedded AI-powered video analytics. It enables video security systems to instantly identify suspicious behavior & abnormal events to prevent potential threats. The VMS provider clients include banking institutions. To avoid theft and unauthorized actions with ATMs, the customer decided to boost its existing solution with the anomaly detection algorithm.
Solution
Anomaly detection is figuring out uncommon activities or observations that can enhance suspicions by being statistically distinct from the rest of the observations. Usually, anomalous data is something that doesn’t fit the common pattern. In other words, data anomaly detection is the identification of outlier values in the data series. Often data anomalies are connected with some kind of rare occurring events. For instance, equipment malfunctioning, medical issues, bank frauds, etc. Based on this relationship, it is possible to determine the points that will be considered anomalies. In our case, the anomaly detection algorithm was supposed to be aimed at identifying suspicious persons heading to the ATM.
Video management software works in its usual mode. As soon as persons with suspicious objects or with abnormal clothing accessories approach ATMs, the built-in analytics identifies this as a deviation from the standard, and an alarm notification is generated for the operator. For instance, the person with the balaclava on his head or the person with strange objects in his hands (f.e. gas balloon, gun, hook, etc.) at X-distance from the ATM is perceived as an anomaly. The operator gets the alarm notification and decides what to do next, skip the event, or call security guards & police. So, the developed algorithm defines anomalies and helps to prevent attacks on ATMs to steal money.
Video analytics software plays a significant role because it helps to respond to unwanted events faster and more efficiently. Such an approach helps banking institutions to secure themselves from fraud, robberies, or other types of illegal actions, reduce rates of crime about them, and increase the confidence of clients in the organization.
Business Impact
Anomaly detection algorithm for identifying suspicious behavior that might lead to robbery and destruction of ATMs was successfully implemented as a part of built-in analytics in VMS. Thanks to the expansion of AI analytics video security provider was able to provide the additional functionality to its client – one of the main German banking institution – and thus to increase security measures with regard to ATMs.
In general, the modern world is the era of state-of-the-art technologies and AI solutions. Object detection, brand recognition, object tracking, face recognition, 3D reconstruction, and lots of other use cases are possible nowadays thanks to the power of technology and progress. The analysis of real-time images & live videos is the reality. Our teams develop custom AI/ML solutions that come in handy by understanding digital pictures and live videos.
Contact our professionals to talk about custom AI/ML solutions that will enhance and power your software.
Anomaly Detection Development Costs
Anomaly detection focuses on the identification of deviations from the standard patterns and behavioral norms. The recognition of unusual activities/anomalous data helps avoid undesirable events in various contexts and industries. Anomaly detection defines irregularities and enables timely intervention, thus contributing to security.
Development Cost
Common anomaly detection development costs: $15,000–$35,000. However, many points affect the cost of development:
- Scope and complexity of the anomaly detection solution. For instance, simple data anomaly detection algorithm development costs less in comparison to the implementation of complex systems with various functionality.
- The quantity, quality, and availability of the data to train & test the model
- Algorithm and ML model selection. The understanding of the data anomalies nature, specific data characteristics, and application requirements influences the selection of the most efficient algorithm and model.
- Hardware and infrastructure. The selection of the hardware should align with the specific business requirements of the solution to ensure efficiency. For the processing of large data volumes arises the need for powerful hardware.
- Integrations. For example, integrations with the existing systems, DBs, etc.
- Testing
Implementation Time
The implementation time of common anomaly detection takes ~1-3 months. In any case, the duration time varies depending on the upper listed factors.
