Screen Capture Sensitive Data Protection in Call Centers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for protecting sensitive information in call center environments are inadequate, as they rely on unreliable agent diligence or application-specific customization, and fail to detect malicious behavior, especially when agents use virtual machines.
Innovation Solution
The system employs optical character recognition to identify sensitive information fields in recorded interactions, blurring sensitive images and analyzing agent behavior to detect potential malicious activity, using video sensors and a server-based analyzer to compare field identifiers with predefined lists and monitor agent interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If blurring is performed upon receiving a specific command from an agent, then sensitive information can be protected, but the system becomes unreliable if the agent is not diligent
Solution Approach 1:
The system automatically detects sensitive information fields through optical character recognition and applies blurring without requiring agent intervention. The recording system self-identifies sensitive data by comparing recognized text against a database of sensitive field identifiers, eliminating dependence on agent diligence while maintaining protection reliability
Solution Approach 2:
The manual mechanical process of agent-initiated blurring is replaced with an automated optical character recognition system. The system uses video sensors to capture screens, performs text recognition, compares against sensitive field databases, and automatically applies blurring - substituting human action with an automated detection and protection mechanism
2Reliability
If blurring is done upon recognizing specific screens appearing on the workstation, then sensitive information can be protected, but the system requires customization for every application and may fail if a new application is executed
Solution Approach 1:
The system uses a universal approach by comparing recognized text fields against a comprehensive database of sensitive information field identifiers rather than requiring application-specific configurations. This text-based identification method works across different applications and screen types, providing universal protection without customization overhead
Solution Approach 2:
The system changes from screen-based identification (which requires knowing specific application screens) to text-content-based identification. By recognizing and comparing text fields against sensitive information databases, the system adapts to any application dynamically without requiring pre-configuration for each application type
3Reliability
If keystroke monitoring is used to detect malicious behavior, then agent behavior can be monitored, but the technique fails to detect virtual keystrokes from virtual machines
Solution Approach 1:
Instead of monitoring keystroke inputs directly, the system captures visual copies of the screen display through video sensors. By analyzing the visual output rather than the input mechanism, the system can detect sensitive information exposure regardless of whether the input came from physical keyboards or virtual machine keystrokes, bypassing the limitation of traditional keystroke monitoring
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach effectively protects sensitive information by pinpoint blurring and analyzing agent behavior, reducing the risk of data breaches and identifying potential malicious activities in real-time, enhancing security and monitoring within call centers.
Implementation Method 1
A plurality of application screens displayed on a workstation of an agent of a call center during the agent's interaction with a customer are recorded via a video sensor
Implementation Method 2
Optical character recognition is performed to identify the identifiers of the fields within the images
Implementation Method 3
At least those portions of those images in the recording that include the data items entered into the fields associated with the sensitive information field identifiers are blurred
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Portions of a customer interaction recording that include sensitive information can be identified by performing character recognition on the images (24) in the recording (23). Field identifiers (21) identified through character recognition are compared to a list of known identifiers (28) of fields for entry of sensitive information. Based on the comparison, field identifiers (21) identified through character recognition can be recognized as sensitive information field identifiers (28). Images (24) that include the sensitive information field identifiers can be blurred in the stored recording (23). Further, agent behavior in relation to the screens with fields for entry of sensitive information can be analyzed for purposes such as identifying potentially malicious agent behavior, an agent being in need of assistance, or a recurring customer issue.