Fraud Detection System Using Browsing Activity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for assessing the trustworthiness of online activities are inefficient due to insufficient information, failure to consider various sources of data, and inability to accurately infer legitimate versus illegitimate behavior, leading to false identification of trustworthy individuals and activities as untrustworthy.
Innovation Solution
A system that analyzes various sources of information, including online browsing activity and transaction data, to infer the legitimacy of transactions by deriving model attributes and applying them to a fraud determination model, allowing for real-time decisions on transaction approval based on calculated probabilities and business considerations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional scoring systems are used to assess trustworthiness, then the system structure is simple, but the accuracy of trustworthiness determination deteriorates due to insufficient information and failure to consider multiple data sources
Solution Approach 1:
The system segments the trustworthiness assessment into multiple independent modules: data collection module (gathering browsing activity, transaction data), data enrichment module (adding contextual information), fraud determination module (applying machine learning models), and decision module (outputting trustworthiness assessment). Each module processes specific data types and feeds into the next, enabling complex analysis while maintaining modular simplicity.
Solution Approach 2:
The patent introduces an intermediary data enrichment module that bridges raw data from multiple sources with the fraud determination model. This intermediary layer processes and contextualizes raw data (browsing patterns, transaction metadata) into meaningful features that the machine learning model can effectively process, enabling accurate trustworthiness assessment without directly exposing the model's complexity to the data collection layer.
2Productivity
If manual intervention is used for trustworthiness assessments, then the system can consider contextual nuances, but productivity deteriorates due to inefficiency and inability to scale
Solution Approach 1:
The system implements self-service through automated machine learning models that independently perform trustworthiness assessments without human intervention. The fraud determination module uses trained models to automatically analyze browsing activity patterns, transaction data, and contextual information, making real-time decisions at scale while maintaining high accuracy through continuous learning from feedback data.
Solution Approach 2:
The system incorporates feedback mechanisms where outcomes of automated assessments (both correct and incorrect) are fed back into the machine learning models for continuous training. This feedback loop enables the system to improve its accuracy over time while maintaining automated high-speed processing, resolving the trade-off between automation speed and assessment precision.
3Reliability
If conventional systems block suspicious activities, then security is improved, but legitimate trustworthy activities are also blocked due to false positives
Solution Approach 1:
The system dynamically adjusts decision parameters and risk thresholds based on contextual analysis of browsing patterns, transaction history, and user behavior profiles. By changing the sensitivity parameters of the fraud detection model adaptively rather than using fixed thresholds, the system can distinguish between legitimate risky behavior and actual fraud, allowing legitimate transactions to proceed while maintaining security.
Solution Approach 2:
The trustworthiness assessment system employs dynamic decision-making that adapts to individual user behavior patterns over time. The machine learning models continuously update user profiles and risk assessments based on observed behavior, enabling the system to distinguish between one-time anomalies and genuine fraud patterns. This dynamic approach reduces false positives by allowing legitimate activities to be recognized and approved as users demonstrate trustworthy behavior patterns.
4Measurement precision
If conventional systems use limited data sources, then the system is simpler to implement, but the ability to accurately infer legitimacy deteriorates
Solution Approach 1:
The system implements a universal data collection framework that aggregates multiple data sources (browsing activity logs, transaction metadata, device information, contextual data) into a unified representation of user behavior. This multi-functional approach allows a single system architecture to process diverse data types from various sources, enabling accurate legitimacy inference without proportionally increasing implementation complexity.
Data Source
AI summary
Various forms of information are utilized in a system and accompanying method for making inferences regarding the trustworthiness of a person performing an online transaction and deciding whether to allow the transaction to have material implications. More specifically, the information relates to the online browsing activity of a user and the online transaction being performed by the user. Further, information regarding certain probable characteristics of the user is determined based on a possible association between the user and one or more known entities. Based on the foregoing information, model attributes are derived and provided as input to a fraud determination model. Using this model and one or more of the attributes, a probability that the transaction is fraudulent is determined.