ML-Based Suspect Session Detection and Activity Halting

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Vishing attacks exploit computer technology to trick users into providing sensitive information or transferring funds through fraudulent phone calls, which are difficult to trace, especially when using prepaid gift cards, as existing detection methods struggle to prevent subsequent fraudulent activities effectively.

Innovation Solution

A computer-based system utilizing a trained detection machine learning model to monitor user activities, identify suspect interaction sessions, and automatically halt future activities based on session interaction parameters and predetermined triggers, preventing further transactions associated with identified fraudulent interactions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If existing detection methods are used to identify fraudulent phone calls, then some fraudulent interactions can be detected, but subsequent fraudulent activities (especially gift card transactions) cannot be effectively prevented

Engineering Contradiction:

Improvefraud detection accuracyVSAvoidprevention of subsequent fraudulent activities

Core Design Contradiction:

ReliabilityVSManufacturing precision

Solution Approach 1:

The system performs preliminary actions by continuously monitoring user activities and pre-identifying suspect interaction sessions before fraudulent transactions occur. The machine learning model analyzes patterns during the suspect session to predict future fraudulent activities, enabling prevention before the actual harm is done.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements feedback by using the outcome of suspect session analysis to automatically halt identified future fraudulent activities. The machine learning model's predictions trigger automatic prevention actions, creating a closed-loop system where detection results directly inform prevention measures.

Inventive Principle:

Principle #23Feedback

2Measurement precision

If manual monitoring and verification of interaction sessions is performed, then fraudulent calls can be identified, but the system cannot respond quickly enough to prevent subsequent transactions

Engineering Contradiction:

Improveidentification of suspect interaction sessionsVSAvoidresponse time to prevent fraudulent transactions

Core Design Contradiction:

Measurement precisionVSSpeed

Solution Approach 1:

The system performs self-service by automatically monitoring activities, identifying suspect sessions, analyzing patterns, and halting future fraudulent activities without human intervention. The machine learning model autonomously makes decisions about which sessions are suspect and triggers automatic prevention actions.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system replaces manual mechanical monitoring with an automated electronic system using machine learning models. The ML analysis substitutes for human analysts, and automatic halting mechanisms replace manual intervention, dramatically increasing response speed while maintaining detection precision.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If comprehensive monitoring of user activities is implemented to detect fraudulent patterns, then detection accuracy improves, but system complexity and resource consumption increase

Engineering Contradiction:

Improvedetection of fraudulent patternsVSAvoidmonitoring system structure

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the monitoring process into distinct functional modules: activity monitoring, suspect session identification, machine learning analysis, and automatic halting. This modular architecture reduces overall system complexity by making each component independent and manageable while maintaining comprehensive detection capabilities.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12105795B2Computer-based systems configured for utilization of a trained detection machine learning model for activity determination and methods of use thereof

Publication Date: 2024.10.01 CAPITAL ONE SERVICES LLC

AI summary

In some embodiments, the present disclosure provides an exemplary method that may include steps of obtaining a permission to monitor a plurality of activities executed within the second computing device; continuously monitoring the plurality of activities executed within the second computing device for a predetermined period of time; receiving an indication of an incoming interaction session within the predetermined period of time; automatically verifying at least one session interaction parameter associated with the incoming interaction session to identify the incoming interaction session as a suspect interaction session; determining when a duration of time associated with the suspect interaction session meets or exceeds a predetermined duration threshold; utilizing a trained detection machine learning model to model a confidence value associated with the suspect interaction session; identifying at least one future activity subsequent to the suspect interaction session; and automatically instructing the first computing device to halt the future activity.