Voice Stream Segmentation for Secure Credit Card Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for secure transmission of sensitive information, such as credit card details, during telephone calls are inadequate, particularly when callers cannot use touch-tone phones or experience poor accuracy with Automatic Speech Recognition due to regional accents or background noise, leading to unacceptable transaction failure rates.
Innovation Solution
A method and system that securely identifies spoken alphanumeric characters by scrambling or partitioning the voice stream before sending it to recognition agents, ensuring that the original sequence cannot be easily reconstructed, thereby enhancing security and accuracy without requiring touch-tone phones or perfect speech recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If ASR software is used to capture spoken digits, then the system can handle callers unable to use touch-tone phones, but the accuracy drops to approximately 68% for complete card numbers
Solution Approach 1:
The patent divides the 19-digit sequence into multiple segments of 3-4 digits each, processed separately through ASR. This segmentation allows the system to handle smaller chunks of information independently, reducing the cumulative error effect and enabling partial success scenarios where some segments are correctly recognized even if others are not.
Solution Approach 2:
The system implements feedback loops where the ASR output is validated against expected patterns (Luhn check digit validation, reasonable digit ranges). When validation fails, the system requests re-speaking of specific segments, allowing iterative improvement of recognition accuracy without requiring perfect initial recognition.
2Loss of information
If the complete voice stream is sent to recognition agents, then all spoken digits can be captured, but security is compromised as agents can access the full sequence
Solution Approach 1:
The voice stream is segmented into multiple voice packets, each containing only a portion of the complete digit sequence. Recognition agents receive and process only their assigned segments, making it computationally infeasible for them to reconstruct the complete card number even if they intercept all packets, as the segmentation keys are distributed and not centrally held.
Solution Approach 2:
The patent introduces an intermediary processing system that acts as a trusted mediator between the caller and recognition agents. This intermediary performs the segmentation, manages the distribution of voice packets to agents, and reconstructs the complete sequence after validation, ensuring that no single agent has access to the complete sensitive information.
3Measurement precision
If verification steps are added to confirm recognized numbers, then accuracy can be improved, but the transaction process becomes slower and more complex
Solution Approach 1:
The system performs preliminary validation of ASR output against known constraints (first digit must be 3, 4, or 5 for PAN; last digit must satisfy Luhn check) before completing the transaction. This preliminary filtering catches obvious errors early, allowing the system to request corrections only when necessary, rather than requiring full verification of every digit sequence.
Solution Approach 2:
The patent changes the validation parameters from requiring perfect 100% digit accuracy to accepting partial matches that satisfy structural constraints. By validating against patterns rather than requiring exact reconstruction, the system can proceed with transactions when the recognized sequence is structurally valid, significantly reducing verification overhead.
4Object-affected harmful factors
If multiple recognition agents are used to process different segments, then security is improved, but the system complexity increases
Solution Approach 1:
The patent designs a universal segmentation and reconstruction framework that can work with any number of recognition agents. The same core algorithms and protocols handle segmentation, distribution, and reconstruction regardless of whether 2, 3, or more agents are involved, allowing the system to scale security by adding agents without proportionally increasing system complexity.
Data Source
AI summary
The present invention relates to a telephone call processing system and a method of processing telephone calls, and in particular to systems and methods that facilitate secure transmission and accurate identification of sensitive information spoken during a call between a caller and an agent such that the agent does not have access to the sensitive information. The methods of the present invention involve outputting scrambled, out-of-sequence and/or partitioned sub-sets of a spoken voice stream to one or more recognition agents, which identify elements of the sensitive information and return the identified elements to a call processor for un-scrambling, re-ordering and/or re-assembly by the call processor to produce the identified sensitive information.