Dynamic Feature Optimization Leveraging Quantum Simulation for Fake Account Detection

Patent metadata

Field	Value
Publication number	US-2025350636-A1
Application number	US-202418595748-A
Country	US
Kind code	A1
Filing date	Mar 5, 2024
Priority date	Mar 5, 2024
Publication date	Nov 13, 2025
Grant date	—

Abstract

Robust systems and methods are disclosed for fake account detection on digital platforms, integrating provenance analysis to scrutinize data origins, ownership, and history, thereby unveiling potential sources of fraudulent activities. They leverage dynamic feature generation, using advanced algorithms to assess user behaviors and interactions, ensuring the model stays attuned to the evolving landscape of cyber threats. Incorporating Quantum-assisted optimization, the method employs Quantum algorithms to expedite feature selection, enhancing detection efficiency. Quantum simulation further refines this process, creating sophisticated verification patterns and analytical techniques to distinguish genuine from fake accounts with higher accuracy. A comprehensive analysis amalgamates provenance data, telemetry, and dynamic features, forming a holistic detection approach. This system optimizes features through Quantum simulation, tailoring them to specific business environments, and deploys them via AI-ML DevOps, streamlining orchestration across various operational settings.

First claim

1 . A method for detecting fake accounts on digital platforms through the integration of Quantum computing and artificial intelligence technologies, the method comprising the steps of:
initiating a data provenance analysis to systematically trace and document origin, ownership, and historical transformations of data collected from digital platforms to identify potential sources of fraudulent activities, wherein the data provenance analysis employs advanced cryptographic techniques to ensure confidentiality and integrity of the data; implementing an algorithmic process for dynamic generation of detection features, wherein said process utilizes real-time analytics of user behaviors, interactions, and engagement patterns to adaptively generate new features that accurately reflect current trends and tactics employed by fraudulent entities, incorporating machine learning models that adjust detection parameters in real-time based on evolving online behaviors; applying Quantum optimization algorithms, specifically leveraging Quantum Annealing and Quantum Approximate Optimization Algorithm (QAOA), to expedite and enhance a selection process of relevant detection features by performing parallel computations and exploring vast combinations of potential features, thereby significantly reducing optimization time while increasing accuracy of feature selection; engaging Quantum simulation techniques to create and refine advanced detection features, wherein the simulation facilitates development of features capable of performing analyses for real-time verification, peer comparison, and behavioral analysis, and enables the simulation of adversarial scenarios to test and improve resilience of a detection system against novel and sophisticated fraud techniques; conducting a comprehensive and integrated analysis that amalgamates insights gained from data provenance, telemetry data, and the dynamically generated and Quantum-optimized features, to form a robust and holistic understanding of a digital environment, ensuring a thorough and nuanced approach to the detection of fake accounts; optimizing the generated features through further Quantum simulations to finalize a set of highly effective detection features, including evaluating the performance of the features in diverse scenarios and refining them based on their efficacy in identifying fake accounts to minimize false positives and negatives, thereby ensuring reliability and trustworthiness of the detection process; deploying the optimized features across various digital platforms through an automated AI-ML DevOps pipeline, which facilitates seamless integration and continuous deployment of detection capabilities, ensuring that the can be dynamically updated in response to emerging threats or changes in the digital environment, including provisions for incremental updates and real-time adjustments without necessitating system downtime; and establishing a real-time verification and analysis component that employs the deployed features to promptly identify and validate potential threats, incorporating a feedback mechanism that allows for the continuous improvement of the detection system based on outcomes of verified threats, thereby maintaining high accuracy and effectiveness in the detection of fake accounts while minimizing impact on legitimate users and operations of the digital platform.

2 . The method of claim 1 , further comprising the step of enhancing the data provenance analysis with artificial intelligence (AI)-based anomaly detection algorithms designed to automatically flag data inconsistencies and potential fraudulent patterns within the provenance data, thereby improving the initial filtering process for identifying suspicious activities and sources of fake accounts.

3 . The method of claim 2 , wherein the dynamic generation of detection features further includes natural language processing (NLP) techniques to analyze textual content within user interactions for sentiment analysis, keyword spotting, and contextual understanding, enhancing identification of nuanced and sophisticated fraudulent activities based on content analysis.

4 . The method of claim 3 , wherein the Quantum optimization algorithms are tailored to prioritize features based on their demonstrated effectiveness in detecting emerging threats, incorporating a machine learning feedback loop that dynamically adjusts feature prioritization based on the latest trends in fraudulent behavior identified through ongoing analysis and Quantum simulation results.

5 . The method of claim 4 , wherein the Quantum simulation for refining advanced detection features includes a component for simulating social network structures and interaction patterns to identify and understand the mechanisms of spread and influence among fake accounts, facilitating the development of features that can detect coordinated inauthentic behavior across the platform.

6 . The method of claim 5 , further including the deployment of an adaptive threshold setting mechanism within the real-time verification and analysis component, wherein the mechanism adjusts sensitivity of the detection system based on current platform dynamics and threat levels, ensuring optimal balance between the detection of fake accounts and the minimization of false positives, thereby maintaining user trust and platform integrity.

7 . A method for detecting fake accounts on digital platforms comprising the steps of:
conducting a comprehensive analysis of data provenance to identify potential fraudulent activities by tracing origin, ownership, and processing history of data, wherein the data provenance analysis includes capturing metadata on data sources, transformations, and intermediate results to preserve lineage of the data; dynamically generating new detection features based on real-time analysis of user behaviors and interactions, wherein the dynamic generation employs algorithms to create features that adapt to evolving tactics used by fraudulent accounts; utilizing Quantum optimization algorithms, including Quantum Annealing, to refine a feature selection process, wherein the Quantum optimization leverages parallel processing capabilities of Quantum computing to explore and optimize multiple feature combinations simultaneously; employing Quantum simulation to further generate advanced detection features, facilitating sophisticated analysis methods including real-time verification, peer comparison verification, content analysis, and behavioral analysis to improve accuracy of fraud detection while minimizing false positives; conducting a comprehensive analysis that integrates provenance, telemetry, and dynamic feature analysis to develop a holistic understanding of fraudulent account activities; optimizing the generated features through Quantum simulation for deployment across various digital environments, wherein the optimization process selects a most effective feature schema for current business conditions and employs AI-ML DevOps for efficient feature deployment; and implementing a real-time verification and analysis component to ensure prompt and accurate validation of detected threats, thereby minimizing impact on legitimate users and swiftly neutralizing potential risks, wherein a real-time component adapts to changing tactics by malicious actors.

8 . The method of claim 7 , wherein the dynamic feature generation process incorporates machine learning models to adjust detection features based on outputs from the data provenance analysis and the real-time analysis of user data, thereby ensuring a detection mechanism remains effective against adaptive strategies by fraudulent accounts.

9 . The method of claim 7 , further comprising a filtering mechanism to prioritize data sources assoc

Assignees

• Bank Of America

Inventors

• Taneja Pushkar
• Adivi Suryanarayana
• Singh Shailendra
• Ryali Venkata Karthik

Classifications

• G06Q20/3823

  combining multiple encryption tools for a transaction · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N10/20

  Models of quantum computing, e.g. quantum circuits or universal quantum computers · CPC title

• G06Q10/06393

  Score-carding, benchmarking or key performance indicator [KPI] analysis · CPC title

• G06N20/00

  Machine learning · CPC title