Machine Learning Input Analysis for Unauthorized Activity Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to effectively differentiate between human and automated inputs, making it difficult to detect unauthorized activity, particularly as malicious actors increasingly use bots to mimic human interactions.
Innovation Solution
A computing platform utilizes machine learning to analyze input patterns, such as cadence and background noise, to distinguish between human and non-human users, and triggers heightened authentication for suspicious inputs, including callbacks and detailed questions to verify identity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional authentication methods are used, then user verification is simple and quick, but the system cannot effectively detect unauthorized activity or differentiate between human and automated inputs
Solution Approach 1:
The system performs preliminary analysis of input patterns during the authentication process itself, examining cadence, timing, and behavioral characteristics before making an authorization decision. This allows the system to detect potential unauthorized activity as part of the normal authentication flow rather than requiring separate complex verification steps.
Solution Approach 2:
The patent introduces machine learning models as intermediary components that analyze input patterns and provide risk assessments between the user's authentication attempt and the system's authorization decision. These models act as mediators that evaluate behavioral characteristics without requiring direct complex interaction with the user.
2Reliability
If the system requires increased authentication credentials for all users, then security is improved, but user experience and convenience deteriorate
Solution Approach 1:
The system applies different levels of authentication scrutiny to different users based on their specific input pattern analysis. Users with human-like cadence and behavioral patterns experience standard simple authentication, while those exhibiting automated or suspicious patterns trigger additional verification steps. This localized application of security measures maintains convenience for legitimate users while enhancing security where needed.
Solution Approach 2:
The system dynamically adjusts authentication requirements based on analyzed parameters such as input cadence, timing patterns, and behavioral characteristics. Rather than applying fixed authentication rules to all users, the system modifies authentication depth based on real-time assessment of input patterns, allowing legitimate users to proceed quickly while flagging suspicious cases for further verification.
Data Source
AI summary
Arrangements for detecting unauthorized activity based on input method analysis and monitoring are provided. In some aspects, identity information associated with a user may be received and be stored. An input may be received from a computing device of the user. An input pattern of the received input may be determined. Using a machine learning model, the input pattern of the received input may be compared to input patterns of humans and input patterns of machines. Based on the comparison, it may be determined whether the user is a human user or a non-human user. Responsive to determining that the user is a non-human user, a request may be transmitted to the user to provide increased authentication credentials. Responsive to determining that the user is a human user, an identity of the user may be verified by comparing the input pattern of the received input to the stored identity information.