Selected work

Representative Papers

Privacy-Preserving Federated Vision Transformer Learning Leveraging Lightweight Homomorphic Encryption in Medical AI

Nov 25, 2025

To address the dual challenges of privacy leakage (e.g., gradient inversion attacks) and high communication overhead in medical federated learning, this paper proposes a privacy-enhancing framework integrating Vision Transformers (ViT) with CKKS homomorphic encryption. The core method replaces conventional encrypted gradient aggregation with homomorphically encrypted CLS tokens from ViT—enabling secure, feature-level aggregation and direct inference directly in the ciphertext domain. This design eliminates gradient reconstruction vulnerabilities while drastically reducing communication costs: per-round overhead drops by 30× to merely 326 KB. Evaluated on a lung cancer histopathological classification task, the framework achieves a global accuracy of 96.12%, with ciphertext-domain inference maintaining 90.02% accuracy. Thus, it simultaneously delivers strong privacy guarantees, substantial communication efficiency gains, and competitive model performance.

0 citationsRead paper

The Differential Meaning of Models: A Framework for Analyzing the Structural Consequences of Semantic Modeling Decisions

Aug 29, 2025

The field of semantic modeling lacks a unified theoretical framework, hindering systematic comparison of semiotic assumptions underlying human meaning construction across diverse models. Method: Grounded in Peircean semiotics, this paper introduces Model Semantics Theory—a formal framework that treats models and their design choices as signs, characterizing their implicit symbolic geometry via latent-variable modeling and relational analysis. Contribution/Results: The framework enables the first cross-model-type comparability—spanning vector spaces, graph neural networks, and logical reasoning systems—revealing how modeling decisions structurally constrain interpretive perspectives. Empirical validation across multiple case studies demonstrates strong explanatory power and scalability. It establishes the first unified analytical paradigm for semantic modeling that integrates theoretical rigor with practical applicability.

0 citationsRead paper

Semiotic Complexity and Its Epistemological Implications for Modeling Culture

Jul 31, 2025

A pervasive methodological insufficiency characterizes contemporary computational humanities: dominant modeling practices routinely reduce semiotically complex cultural texts—such as language and narrative—to semiotically simple formal structures, inducing semantic mistranslation, interpretive opacity, and epistemic bias. This paper introduces the novel concept of “semiotic complexity,” grounded in interdisciplinary critical analysis spanning semiotics, linguistics, and computational modeling, to systematically expose the mechanisms of semantic mistranslation arising during model evaluation. Building on this, we propose a theoretically grounded framework for translating cultural texts into mathematical models and articulate four principles for identifying and mitigating semiotic translation errors. The work advances computational humanities from technical application toward methodological reflexivity, substantially enhancing model interpretability, explanatory rigor, and theoretical transparency. (149 words)

0 citationsRead paper

AI-Driven Secure Data Sharing: A Trustworthy and Privacy-Preserving Approach

Jan 26, 2025

To address the privacy-accuracy-efficiency trade-off in encrypted-domain deep learning inference for sensitive data sharing, this paper proposes a learnable, key-customized block-wise pixel scrambling encryption method, tightly integrated with Vision Transformers (ViTs). This integration enables, for the first time, end-to-end secure ViT inference over dynamically encrypted inputs. The approach performs high-accuracy, low-overhead model inference directly in the ciphertext domain—without decryption—thereby simultaneously ensuring strong privacy guarantees, computational efficiency, and robustness. Adversarial robust training is further incorporated to enhance security against malicious perturbations. Evaluated on MRI brain tumor and lung/colon cancer pathology datasets, the method achieves 94% classification accuracy and demonstrates strong resilience against diverse adversarial attacks. These results validate its reliability and practicality in real-world, high-sensitivity medical applications.

0 citationsRead paper

Recent publications

Latest Papers

Privacy-Preserving Federated Vision Transformer Learning Leveraging Lightweight Homomorphic Encryption in Medical AI

Nov 25, 2025

To address the dual challenges of privacy leakage (e.g., gradient inversion attacks) and high communication overhead in medical federated learning, this paper proposes a privacy-enhancing framework integrating Vision Transformers (ViT) with CKKS homomorphic encryption. The core method replaces conventional encrypted gradient aggregation with homomorphically encrypted CLS tokens from ViT—enabling secure, feature-level aggregation and direct inference directly in the ciphertext domain. This design eliminates gradient reconstruction vulnerabilities while drastically reducing communication costs: per-round overhead drops by 30× to merely 326 KB. Evaluated on a lung cancer histopathological classification task, the framework achieves a global accuracy of 96.12%, with ciphertext-domain inference maintaining 90.02% accuracy. Thus, it simultaneously delivers strong privacy guarantees, substantial communication efficiency gains, and competitive model performance.

0 citationsRead paper

The Differential Meaning of Models: A Framework for Analyzing the Structural Consequences of Semantic Modeling Decisions

Aug 29, 2025

The field of semantic modeling lacks a unified theoretical framework, hindering systematic comparison of semiotic assumptions underlying human meaning construction across diverse models. Method: Grounded in Peircean semiotics, this paper introduces Model Semantics Theory—a formal framework that treats models and their design choices as signs, characterizing their implicit symbolic geometry via latent-variable modeling and relational analysis. Contribution/Results: The framework enables the first cross-model-type comparability—spanning vector spaces, graph neural networks, and logical reasoning systems—revealing how modeling decisions structurally constrain interpretive perspectives. Empirical validation across multiple case studies demonstrates strong explanatory power and scalability. It establishes the first unified analytical paradigm for semantic modeling that integrates theoretical rigor with practical applicability.

0 citationsRead paper

Semiotic Complexity and Its Epistemological Implications for Modeling Culture

Jul 31, 2025

A pervasive methodological insufficiency characterizes contemporary computational humanities: dominant modeling practices routinely reduce semiotically complex cultural texts—such as language and narrative—to semiotically simple formal structures, inducing semantic mistranslation, interpretive opacity, and epistemic bias. This paper introduces the novel concept of “semiotic complexity,” grounded in interdisciplinary critical analysis spanning semiotics, linguistics, and computational modeling, to systematically expose the mechanisms of semantic mistranslation arising during model evaluation. Building on this, we propose a theoretically grounded framework for translating cultural texts into mathematical models and articulate four principles for identifying and mitigating semiotic translation errors. The work advances computational humanities from technical application toward methodological reflexivity, substantially enhancing model interpretability, explanatory rigor, and theoretical transparency. (149 words)

0 citationsRead paper

AI-Driven Secure Data Sharing: A Trustworthy and Privacy-Preserving Approach

Jan 26, 2025

To address the privacy-accuracy-efficiency trade-off in encrypted-domain deep learning inference for sensitive data sharing, this paper proposes a learnable, key-customized block-wise pixel scrambling encryption method, tightly integrated with Vision Transformers (ViTs). This integration enables, for the first time, end-to-end secure ViT inference over dynamically encrypted inputs. The approach performs high-accuracy, low-overhead model inference directly in the ciphertext domain—without decryption—thereby simultaneously ensuring strong privacy guarantees, computational efficiency, and robustness. Adversarial robust training is further incorporated to enhance security against malicious perturbations. Evaluated on MRI brain tumor and lung/colon cancer pathology datasets, the method achieves 94% classification accuracy and demonstrates strong resilience against diverse adversarial attacks. These results validate its reliability and practicality in real-world, high-sensitivity medical applications.

0 citationsRead paper