ATM Camera Image Blending for Skimming Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current self-service terminals, such as ATMs, face challenges in reliably detecting manipulation attempts without additional sensors, as existing camera surveillance systems struggle to differentiate between legitimate and malicious activities, particularly with skimming devices that mimic original controls.
Innovation Solution
A self-service terminal equipped with multiple cameras that generate image data based on predeterminable criteria, such as time intervals and lighting conditions, with a data processing unit that combines and preprocesses this data using methods like segmentation, edge detection, and exposure blending to create high-quality result images for manipulation detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If camera surveillance systems are used to monitor the control panel and user area, then manipulation attempts can be detected, but the system cannot reliably differentiate between legitimate and malicious activities due to skimming devices that mimic original controls
Solution Approach 1:
The control panel is divided into multiple detection zones (keyboard area, card reader area, display area) with dedicated cameras positioned to capture each zone specifically. This segmentation allows the system to focus on specific areas where manipulation is most likely to occur and to differentiate between normal operation and malicious activity in each zone.
Solution Approach 2:
Different camera positions and detection parameters are applied to different zones of the control panel based on their specific characteristics and manipulation risks. The keyboard area has one detection configuration while the card reader area has another, allowing optimized detection for each local region's specific manipulation patterns.
2Measurement precision
If multiple cameras are mounted close to the control panel to capture specific elements, then detection coverage is improved, but the device complexity increases
Solution Approach 1:
The cameras are integrated into the existing terminal housing structure, serving both as security monitoring devices and as part of the terminal's overall design. The same camera system detects multiple types of manipulation attempts (skimming devices, hidden cameras, unauthorized keypads) while maintaining a unified structural implementation.
Solution Approach 2:
Cameras are positioned within or integrated with the terminal housing and control panel structure, nesting the detection system within the existing device architecture. This reduces the need for separate external mounting structures and simplifies the overall system integration.
3Measurement precision
If image data from multiple individual images is combined through preprocessing, then high-quality result images are produced for manipulation detection, but processing time and computational resources increase
Solution Approach 1:
Image preprocessing operations (alignment, normalization, enhancement) are performed in advance before the actual manipulation detection algorithm is applied. This preliminary processing prepares the image data in optimal formats, reducing the computational burden during the critical detection phase and enabling faster processing.
Solution Approach 2:
The image processing is divided into separate stages: individual image capture, preprocessing (alignment, enhancement), and final manipulation detection. This segmentation allows optimization of each stage independently and enables parallel processing of multiple images, reducing overall processing time while maintaining high quality.
Data Source
Figure 1~2
Figure 3a~3d
Figure 4a1~4c
AI summary
The invention proposes an automated teller machine having at least one camera to detect manipulation attempts, said camera capturing images of one or more elements arranged in the control panel, such as the keyboard, money-dispensing compartment, and card entry slot, for example, and producing image data from a plurality of individual image recordings (F1, F2, F3). The one or more cameras are connected to a data processing unit that preprocesses the produced image data (individual image data) into a resulting image (R). The preprocessed image data of the resulting image (R) can be computed, for example, from the individual images (F1, F2, F3) by means of exposure blending and represents a very good database for data analysis to detect manipulation.