Phased Voice Call Fraud Detection With Real-Time User Warnings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to effectively detect fraudulent calls due to the dynamic nature of scammer tactics, including the use of recycled phone numbers and AI-generated voices, leading to increased inconvenience and potential financial and security risks for consumers.
Innovation Solution
A system that segments ongoing voice calls into discrete segments, analyzes each segment for fraudulent patterns using machine learning, and provides real-time warnings to users before sensitive information is disclosed, incorporating large language models and deep neural networks to identify deceptive conversation phases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional call detection systems are used, then the system complexity remains low, but the detection accuracy and reliability are insufficient due to evolving scammer tactics
Solution Approach 1:
The patent segments the call detection process into multiple phases: initial call phase, information gathering phase, and closing phase. Each phase is analyzed separately with phase-specific features extracted and evaluated. This segmentation allows the system to handle different scammer tactics in each phase independently, improving detection accuracy without requiring a monolithic complex system.
Solution Approach 2:
The system performs preliminary analysis during the initial call phase by extracting speaker characteristics, background noise profiles, and conversation patterns before the scammer reveals their intent. This preliminary action establishes a baseline that enhances subsequent detection accuracy while keeping the overall system architecture manageable.
2Loss of time
If real-time analysis of ongoing calls is performed, then the detection timeliness is improved, but the processing time and computational resources increase
Solution Approach 1:
The system performs analysis at periodic intervals corresponding to phase transitions in the conversation rather than continuously analyzing every moment of the call. It extracts features at specific phases (initial, information gathering, closing) and updates fraud scores periodically, reducing computational load while maintaining timely detection capability.
Solution Approach 2:
The system applies partial analysis by focusing computational resources on extracting only the most discriminative features for each phase rather than analyzing all possible conversation attributes continuously. This selective partial action reduces processing requirements while maintaining effective detection timing.
3Measurement precision
If multiple analysis phases are implemented, then the detection precision is improved, but the system complexity and processing overhead increase
Solution Approach 1:
The patent divides the detection process into distinct phases (initial call, information gathering, closing) with specific analysis objectives for each phase. This segmentation improves precision by tailoring analysis to phase-specific characteristics while managing complexity through modular phase-handling routines that can be implemented sequentially.
Solution Approach 2:
The system dynamically adjusts its analysis depth and feature extraction focus based on the current call phase. During the initial phase, it focuses on speaker characteristics; during information gathering, it analyzes conversation patterns; during closing, it evaluates disclosure risk. This dynamic adaptation improves precision without requiring all analysis components to operate simultaneously, thus managing complexity.
4Reliability
If continuous monitoring of calls is performed, then the detection reliability is improved, but the user inconvenience and false alarm risk increase
Solution Approach 1:
The system monitors calls periodically at phase transition points rather than continuously, updating fraud scores and providing warnings only when phase-specific criteria are met. This periodic monitoring maintains reliable detection capability while minimizing user inconvenience by reducing the frequency of system interventions and false alarms.
Solution Approach 2:
The system provides phased feedback to users based on detection confidence levels and phase-specific risk assessments. Rather than constant warnings, it provides targeted feedback when fraud indicators are detected in specific phases, improving reliability while maintaining ease of operation through context-aware warning generation.
Data Source
AI summary
A system and method for detecting fraudulent call activity include segmenting an ongoing voice call between a user and a second party into discrete segments while the call is in progress. The method analyzes respective discrete segments and assigning per-segment weighted fraud scores, where each weighted fraud score accounts for the weighted fraud score of a previous segment. Based on these per-segment weighted fraud scores, the method determines that the voice call is likely a fraudulent call. After making this determination, the method provides a human-perceptible warning to the user before the user discloses sensitive user data.