A network attribution assessment method based on geopolitical bias correction and bayesian uncertainty modeling

By correcting geopolitical bias and using Bayesian uncertainty modeling, the challenges of intelligence position bias and false flag operation identification in network attribution were solved, enabling highly reliable network attack attribution and improving the system's robustness and decision-making transparency.

CN122394873APending Publication Date: 2026-07-14GUANGZHOU UNIVERSITY HUANGPU RESEARCH INSTITUTE +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGZHOU UNIVERSITY HUANGPU RESEARCH INSTITUTE
Filing Date
2026-04-20
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing network attribution technologies suffer from intelligence bias, difficulty in identifying false flag actions, and inability to quantify inference risks, leading to misjudgments and poor robustness.

Method used

By employing a method based on geopolitical bias correction and Bayesian uncertainty modeling, this approach quantifies and removes biases from intelligence sources by constructing an international relations matrix, introducing a nonlinear smoothing damping function, and using a Bayesian deep learning framework. Furthermore, it identifies false flag operations and unknown attack patterns through multi-dimensional uncertainty assessment.

Benefits of technology

It significantly improves the anti-interference capability of the network attribution system, enhances the ability to identify false flag actions, provides multi-dimensional credible decision-making references, and improves the transparency and decision-making reference value of attribution conclusions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122394873A_ABST
    Figure CN122394873A_ABST
Patent Text Reader

Abstract

This invention discloses a network attribution evaluation method based on geopolitical bias correction and Bayesian uncertainty modeling, relating to the intersection of cyberspace security and artificial intelligence. This method addresses existing network attribution techniques' problems of intelligence source bias, difficulty in identifying adversarial interference from "false flags," and unquantifiable inference risks, proposing a complete solution. The main steps include: constructing multi-source heterogeneous data feature engineering and international relations spatial modeling to generate "technology-politics" two-layer triples and association pairs; correcting the prior reputation weights of input-side technical evidence based on a dynamic international relations matrix and a smoothing damping function to remove political noise; and constructing a Bayesian deep learning framework based on MC Dropout to decouple and quantify accidental and cognitive uncertainties through Monte Carlo sampling, outputting attribution conclusions with confidence intervals. This invention achieves semantic fusion of micro-level technical features and macro-level geopolitical background, significantly improving the objectivity, robustness, and decision-making transparency of network attribution.
Need to check novelty before this filing date? Find Prior Art