Geometric Function Atlas: certified computing for geometric function theory in Python
本文介绍了一个开源Python包，用于解决几何函数理论中的极值问题，通过计算泰勒系数、Fekete-Szegő常数等，并提供不同级别的验证方法。
0 citationsRead paper
Selected work
Representative Papers
The Geometric Function Atlas: A Software System for Radius and Coefficient Problems
This study addresses computational inconsistencies and literature errors in geometric function theory by developing an atlas software system featuring directed radius graph representations and verifiable exact certificate mechanisms. Leveraging analytic control, Ma-Minda extremal constructions, and symbolic computation, the project achieves standardization and automated solving for function classes. It establishes nineteen exact sharp radii, corrects a 7.45% error in the sine-to-modified-sigmoid radius, verifies 216 Fekete-Szegő coefficient values, and compiles 702 comparative datasets. These contributions effectively unify multi-named function class systems and rectify existing constant errors, providing a rigorous computational framework for resolving longstanding discrepancies in radius and coefficient calculations within the field.
0 citationsRead paper
Anchors that Don't Lift: Understanding Supply Chain Driven Kernel Lock-In and Governance-Mediated Mitigation Strategies in SOHO Devices
This study addresses the widespread security risks in SOHO devices stemming from outdated Linux kernels, whose real-world impact and update barriers in highly customized environments remain poorly understood. The authors propose a high-precision template-matching method to detect genuine kernel vulnerabilities within GPL-released source code from 306 devices and, for the first time, conduct a large-scale supply chain traceability analysis. Their investigation reveals a pervasive “kernel lock-in” phenomenon caused by SoC vendors’ SDKs, which propagates vulnerability debt downstream. All five major SoC vendors examined rely on kernel SDKs that have been end-of-life for over a year. Regulatory compliance proves insufficient to drive updates; only vendors actively collaborating with the open-source community have achieved meaningful mitigation, underscoring the critical role of community governance in enabling effective kernel upgrades.
0 citationsRead paper
SecureFed: A Two-Phase Framework for Detecting Malicious Clients in Federated Learning
Federated learning is vulnerable to malicious client poisoning attacks, degrading model performance and compromising privacy-preserving guarantees. To address this, we propose a two-stage adaptive defense framework that requires neither a trusted server nor prior knowledge of clients. In the first stage, anomaly detection identifies compromised model updates via PCA-based dimensionality reduction and synthetic data validation. In the second stage, we introduce a novel “learning zone” dynamic weight routing mechanism that jointly leverages gradient magnitude and contribution scoring to suppress low-value (potentially malicious) gradient regions. Evaluated across multiple benchmark datasets, our method achieves over 92% poisoning attack mitigation success rate while incurring less than 0.8% global model accuracy degradation. The framework significantly enhances both robustness and practical deployability without introducing substantial computational overhead or trust assumptions.
0 citationsRead paper
AndroIDS : Android-based Intrusion Detection System using Federated Learning
To address escalating privacy leakage risks and collaborative intrusion detection challenges in Android-based mobile IoT devices (e.g., smart home appliances, drones), this paper proposes a privacy-preserving federated intrusion detection system. Methodologically, it introduces the first federated learning framework specifically tailored to Android system-call trace modeling, employing an LSTM/CNN hybrid architecture to process lightweight, personalized, and non-IID behavioral sequences—enabling localized model training while maintaining global threat awareness. Key contributions include: (1) eliminating raw data transmission to preserve user privacy; (2) achieving high robustness under non-IID data distributions; and (3) demonstrating strong empirical performance—96.46% accuracy and 89% F1-score under IID conditions, and 92.87% accuracy and 86% F1-score under non-IID conditions—significantly outperforming centralized baselines and validating practical deployability.
0 citationsRead paper
Recent publications
Latest Papers
Geometric Function Atlas: certified computing for geometric function theory in Python
本文介绍了一个开源Python包，用于解决几何函数理论中的极值问题，通过计算泰勒系数、Fekete-Szegő常数等，并提供不同级别的验证方法。
0 citationsRead paper
The Geometric Function Atlas: A Software System for Radius and Coefficient Problems
This study addresses computational inconsistencies and literature errors in geometric function theory by developing an atlas software system featuring directed radius graph representations and verifiable exact certificate mechanisms. Leveraging analytic control, Ma-Minda extremal constructions, and symbolic computation, the project achieves standardization and automated solving for function classes. It establishes nineteen exact sharp radii, corrects a 7.45% error in the sine-to-modified-sigmoid radius, verifies 216 Fekete-Szegő coefficient values, and compiles 702 comparative datasets. These contributions effectively unify multi-named function class systems and rectify existing constant errors, providing a rigorous computational framework for resolving longstanding discrepancies in radius and coefficient calculations within the field.
0 citationsRead paper
Anchors that Don't Lift: Understanding Supply Chain Driven Kernel Lock-In and Governance-Mediated Mitigation Strategies in SOHO Devices
This study addresses the widespread security risks in SOHO devices stemming from outdated Linux kernels, whose real-world impact and update barriers in highly customized environments remain poorly understood. The authors propose a high-precision template-matching method to detect genuine kernel vulnerabilities within GPL-released source code from 306 devices and, for the first time, conduct a large-scale supply chain traceability analysis. Their investigation reveals a pervasive “kernel lock-in” phenomenon caused by SoC vendors’ SDKs, which propagates vulnerability debt downstream. All five major SoC vendors examined rely on kernel SDKs that have been end-of-life for over a year. Regulatory compliance proves insufficient to drive updates; only vendors actively collaborating with the open-source community have achieved meaningful mitigation, underscoring the critical role of community governance in enabling effective kernel upgrades.
0 citationsRead paper
SecureFed: A Two-Phase Framework for Detecting Malicious Clients in Federated Learning
Federated learning is vulnerable to malicious client poisoning attacks, degrading model performance and compromising privacy-preserving guarantees. To address this, we propose a two-stage adaptive defense framework that requires neither a trusted server nor prior knowledge of clients. In the first stage, anomaly detection identifies compromised model updates via PCA-based dimensionality reduction and synthetic data validation. In the second stage, we introduce a novel “learning zone” dynamic weight routing mechanism that jointly leverages gradient magnitude and contribution scoring to suppress low-value (potentially malicious) gradient regions. Evaluated across multiple benchmark datasets, our method achieves over 92% poisoning attack mitigation success rate while incurring less than 0.8% global model accuracy degradation. The framework significantly enhances both robustness and practical deployability without introducing substantial computational overhead or trust assumptions.
0 citationsRead paper
AndroIDS : Android-based Intrusion Detection System using Federated Learning
To address escalating privacy leakage risks and collaborative intrusion detection challenges in Android-based mobile IoT devices (e.g., smart home appliances, drones), this paper proposes a privacy-preserving federated intrusion detection system. Methodologically, it introduces the first federated learning framework specifically tailored to Android system-call trace modeling, employing an LSTM/CNN hybrid architecture to process lightweight, personalized, and non-IID behavioral sequences—enabling localized model training while maintaining global threat awareness. Key contributions include: (1) eliminating raw data transmission to preserve user privacy; (2) achieving high robustness under non-IID data distributions; and (3) demonstrating strong empirical performance—96.46% accuracy and 89% F1-score under IID conditions, and 92.87% accuracy and 86% F1-score under non-IID conditions—significantly outperforming centralized baselines and validating practical deployability.
0 citationsRead paper