Selected work

Representative Papers

Privacy-Preserving Credit Card Approval Using Homomorphic SVM: Toward Secure Inference in FinTech Applications

May 09, 2025

In sensitive financial technology applications—such as credit card approval—the high computational overhead of fully homomorphic encryption (FHE) hinders practical deployment, making it challenging to simultaneously achieve strong privacy guarantees and computational efficiency. Method: This paper proposes the first CKKS-based homomorphic encryption framework for soft-margin support vector machines (SVMs). It innovatively incorporates a hybrid kernel function to model nonlinear credit features in ciphertext space and introduces a ciphertext-domain adaptive thresholding mechanism to enhance classification robustness. Contribution/Results: Evaluated on the Credit Card Approval dataset, the method achieves accuracy comparable to plaintext SVM (within ±0.3%), while reducing inference latency by 42%. It is the first approach to achieve a rigorous three-way trade-off among privacy (via FHE), prediction accuracy, and computational efficiency. The framework provides a verifiable, practically deployable solution for privacy-preserving, cloud-based financial decision-making.

0 citationsRead paper

A Novel Feature-Aware Chaotic Image Encryption Scheme For Data Security and Privacy in IoT and Edge Networks

May 01, 2025

To address the low encryption efficiency and weak security of image encryption on resource-constrained devices in IoT and edge networks, this paper proposes a lightweight feature-driven chaotic image encryption scheme. Methodologically, it introduces Feature-Aware Pixel Segmentation (FAPS), integrated with SHA-256–driven dynamic chaotic permutation and seed-matrix-guided chaotic XOR diffusion, forming a three-stage synergistic encryption mechanism. Key innovations include: (1) FAPS, the first to incorporate edge-strength features into pixel grouping, significantly enhancing the specificity and effectiveness of scrambling; and (2) real-time coupling of dynamic keys with chaotic system parameters, substantially improving resistance against differential attacks. Experimental results demonstrate near-zero inter-pixel correlation, information entropy exceeding 7.998, real-time encryption/decryption throughput, and 62% lower power consumption compared to AES—making the scheme highly suitable for low-power edge devices.

0 citationsRead paper

Efficient IoT Intrusion Detection with an Improved Attention-Based CNN-BiLSTM Architecture

Mar 25, 2025

To address the growing challenge of botnet attack detection in IoT systems, this paper proposes a lightweight and efficient intrusion detection method. Methodologically, it introduces a novel CNN-BiLSTM hybrid architecture enhanced with an adaptive attention mechanism, jointly enabling local traffic pattern recognition and long-term temporal dependency modeling while emphasizing discriminative features. Evaluated on the N-BaIoT dataset, the model achieves 99% detection accuracy with high precision and recall; Matthews Correlation Coefficient and Cohen’s Kappa both approach near-ideal values (≈0.98), significantly outperforming existing lightweight models. Key contributions include: (1) the first attention-enhanced, time-series–convolutional joint model designed specifically for IoT edge deployment; (2) a comprehensive multi-dimensional robustness evaluation framework; and (3) state-of-the-art detection performance under low computational overhead, demonstrating strong practical deployability.

0 citationsRead paper

Cross-Platform Benchmarking of the FHE Libraries: Novel Insights into SEAL and Openfhe

Mar 14, 2025

Evaluating performance, memory overhead, and cryptographic scheme support across mainstream fully homomorphic encryption (FHE) libraries remains challenging due to platform-specific optimizations and inconsistent benchmarking methodologies. Method: We conduct a systematic cross-platform evaluation of SEAL and OpenFHE on Linux and Windows, covering the BGV and CKKS schemes. We introduce the first open, cross-platform FHE benchmarking framework (C++-based), employing standardized HE parameters and realistic workloads to isolate and quantify OS-level impacts on homomorphic computation efficiency. Contribution/Results: Our evaluation reveals that Linux delivers 22% higher average throughput than Windows. OpenFHE outperforms SEAL by 1.8–3.4× in CKKS/BGV throughput while reducing memory consumption by 37%, and natively supports multiple cryptographic schemes. These findings provide reproducible, empirical guidance for selecting and deploying FHE libraries in privacy-preserving computing systems.

0 citationsRead paper

X-Cross: Image Encryption Featuring Novel Dual-Layer Block Permutation and Dynamic Substitution Techniques

Mar 13, 2025

To address security threats to digital images during transmission and storage, this paper proposes the X-Cross image encryption algorithm. The method integrates a two-layer block permutation—comprising bit-level IBT permutation and a novel X-cross pixel permutation—with a key-driven dynamic S-box confusion mechanism. A strong nonlinear 2D chaotic map generates dynamic keys, while a Bit-Extraction mechanism enables key-sensitive bit-level rearrangement, effectively eliminating statistical redundancy and structural features. Experimental results demonstrate an information entropy of 7.9993, adjacent-pixel correlation coefficients below 0.002, and robust resistance against differential and statistical attacks—achieving ideal security metrics. The core contributions include: (i) the first two-layer block permutation architecture; (ii) the X-cross pixel permutation algorithm; and (iii) a synergistic chaotic–bit-extraction key generation mechanism.

0 citationsRead paper

Recent publications

Latest Papers

Privacy-Preserving Credit Card Approval Using Homomorphic SVM: Toward Secure Inference in FinTech Applications

May 09, 2025

In sensitive financial technology applications—such as credit card approval—the high computational overhead of fully homomorphic encryption (FHE) hinders practical deployment, making it challenging to simultaneously achieve strong privacy guarantees and computational efficiency. Method: This paper proposes the first CKKS-based homomorphic encryption framework for soft-margin support vector machines (SVMs). It innovatively incorporates a hybrid kernel function to model nonlinear credit features in ciphertext space and introduces a ciphertext-domain adaptive thresholding mechanism to enhance classification robustness. Contribution/Results: Evaluated on the Credit Card Approval dataset, the method achieves accuracy comparable to plaintext SVM (within ±0.3%), while reducing inference latency by 42%. It is the first approach to achieve a rigorous three-way trade-off among privacy (via FHE), prediction accuracy, and computational efficiency. The framework provides a verifiable, practically deployable solution for privacy-preserving, cloud-based financial decision-making.

0 citationsRead paper

A Novel Feature-Aware Chaotic Image Encryption Scheme For Data Security and Privacy in IoT and Edge Networks

May 01, 2025

To address the low encryption efficiency and weak security of image encryption on resource-constrained devices in IoT and edge networks, this paper proposes a lightweight feature-driven chaotic image encryption scheme. Methodologically, it introduces Feature-Aware Pixel Segmentation (FAPS), integrated with SHA-256–driven dynamic chaotic permutation and seed-matrix-guided chaotic XOR diffusion, forming a three-stage synergistic encryption mechanism. Key innovations include: (1) FAPS, the first to incorporate edge-strength features into pixel grouping, significantly enhancing the specificity and effectiveness of scrambling; and (2) real-time coupling of dynamic keys with chaotic system parameters, substantially improving resistance against differential attacks. Experimental results demonstrate near-zero inter-pixel correlation, information entropy exceeding 7.998, real-time encryption/decryption throughput, and 62% lower power consumption compared to AES—making the scheme highly suitable for low-power edge devices.

0 citationsRead paper

Efficient IoT Intrusion Detection with an Improved Attention-Based CNN-BiLSTM Architecture

Mar 25, 2025

To address the growing challenge of botnet attack detection in IoT systems, this paper proposes a lightweight and efficient intrusion detection method. Methodologically, it introduces a novel CNN-BiLSTM hybrid architecture enhanced with an adaptive attention mechanism, jointly enabling local traffic pattern recognition and long-term temporal dependency modeling while emphasizing discriminative features. Evaluated on the N-BaIoT dataset, the model achieves 99% detection accuracy with high precision and recall; Matthews Correlation Coefficient and Cohen’s Kappa both approach near-ideal values (≈0.98), significantly outperforming existing lightweight models. Key contributions include: (1) the first attention-enhanced, time-series–convolutional joint model designed specifically for IoT edge deployment; (2) a comprehensive multi-dimensional robustness evaluation framework; and (3) state-of-the-art detection performance under low computational overhead, demonstrating strong practical deployability.

0 citationsRead paper

Cross-Platform Benchmarking of the FHE Libraries: Novel Insights into SEAL and Openfhe

Mar 14, 2025

Evaluating performance, memory overhead, and cryptographic scheme support across mainstream fully homomorphic encryption (FHE) libraries remains challenging due to platform-specific optimizations and inconsistent benchmarking methodologies. Method: We conduct a systematic cross-platform evaluation of SEAL and OpenFHE on Linux and Windows, covering the BGV and CKKS schemes. We introduce the first open, cross-platform FHE benchmarking framework (C++-based), employing standardized HE parameters and realistic workloads to isolate and quantify OS-level impacts on homomorphic computation efficiency. Contribution/Results: Our evaluation reveals that Linux delivers 22% higher average throughput than Windows. OpenFHE outperforms SEAL by 1.8–3.4× in CKKS/BGV throughput while reducing memory consumption by 37%, and natively supports multiple cryptographic schemes. These findings provide reproducible, empirical guidance for selecting and deploying FHE libraries in privacy-preserving computing systems.

0 citationsRead paper

X-Cross: Image Encryption Featuring Novel Dual-Layer Block Permutation and Dynamic Substitution Techniques

Mar 13, 2025

To address security threats to digital images during transmission and storage, this paper proposes the X-Cross image encryption algorithm. The method integrates a two-layer block permutation—comprising bit-level IBT permutation and a novel X-cross pixel permutation—with a key-driven dynamic S-box confusion mechanism. A strong nonlinear 2D chaotic map generates dynamic keys, while a Bit-Extraction mechanism enables key-sensitive bit-level rearrangement, effectively eliminating statistical redundancy and structural features. Experimental results demonstrate an information entropy of 7.9993, adjacent-pixel correlation coefficients below 0.002, and robust resistance against differential and statistical attacks—achieving ideal security metrics. The core contributions include: (i) the first two-layer block permutation architecture; (ii) the X-cross pixel permutation algorithm; and (iii) a synergistic chaotic–bit-extraction key generation mechanism.

0 citationsRead paper