Hoh Peter In
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hoh Peter In has a total of 79 co-authors. Here are the top co-authors by number of shared publications:
- Mpyana Mwamba Merlec
Department of Computer Science and Engineering, Korea University, Seoul 02841, Republic of Korea
12 shared publications - Mainul Islam
Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX 77843, USA
7 shared publications - Min-Gwan Kim
Korea University College of Informatics Department of Computer Science and Engineering Seoul Republic of Korea
4 shared publications - Chul-Hyun Cho
Department of Psychiatry, Korea University College of Medicine, Seoul, Korea
3 shared publications - Dongwon Jeong
Software Convergence Engineering Department, Kunsan National University, Gunsan 54150, Korea
3 shared publications - Heon-Jeong Lee
Chronobiology Institute, Korea University, Seoul, Korea
3 shared publications
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7464-7468
Beneath the Noise: Uncovering and Distilling the Encoding Bottleneck of Quantum Models
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7464-7468
Open AccessBeneath the Noise: Uncovering and Distilling the Encoding Bottleneck of Quantum Models
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 53432-53446
Verifiable Quantum Machine Learning Inference Consistency for Classical Clients in Noisy Quantum Clouds
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 53432-53446
Verifiable Quantum Machine Learning Inference Consistency for Classical Clients in Noisy Quantum Clouds
- Open Access
- 27 August 2026
- Published by Elsevier in International Journal of Medical Informatics
- Vol. 222
Exposing and addressing the clinical safety gap in ESI-free emergency department triage: Imbalance-aware, interpretable deep learning
- 27 August 2026
- Published by Elsevier in International Journal of Medical Informatics
- Vol. 222
Open AccessExposing and addressing the clinical safety gap in ESI-free emergency department triage: Imbalance-aware, interpretable deep learning
- Open Access
- 21 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Vehicular Technology
- Vol. 7, 1534-1547
Federated Aggregation Scheme With Authentication for Securing IoT-Enabled Social Internet of Vehicles: Enhancing Scalability and Loss Optimization
In the Social Internet of Vehicles (SIoV) based on Digital Twin (DT), which incorporates the social networking concepts from the Social Internet of Things (SIoT), Federated Learning (FL) enables distributed vehicular clients to collaboratively learn machine learning models while preserving data privacy and minimizing communication overhead. However, susceptible model parameters render FL vulnerable to inference and impersonation attacks, necessitating strong and lightweight authentication protocols. This paper introduces a federated learning-based collaborative authentication protocol that incorporates anonymous mutual authentication and dynamic key agreement between vehicles and infrastructure nodes, enabling privacy-preserving data sharing while eliminating reliance on centralized entities. The protocol is evaluated using three aggregation strategies: FedAvg, FedProx, and FedOpt under varying client participation levels and epoch configurations. Experimental results demonstrate that FedOpt achieves superior convergence and generalization, reducing average loss by approximately 7.0% compared to FedAvg and 6.5% compared to FedProx, particularly in heterogeneous environments. Security analysis under the stochastic predictive machine model further validates robustness against adversarial threats. Overall, the proposed protocol enhances scalability, reduces propagation delay, and ensures efficient, privacy-preserving authentication in dynamic SIoV networks.
- 21 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Vehicular Technology
- Vol. 7, 1534-1547
Open AccessFederated Aggregation Scheme With Authentication for Securing IoT-Enabled Social Internet of Vehicles: Enhancing Scalability and Loss Optimization
In the Social Internet of Vehicles (SIoV) based on Digital Twin (DT), which incorporates the social networking concepts from the Social Internet of Things (SIoT), Federated Learning (FL) enables distributed vehicular clients to collaboratively learn machine learning models while preserving data privacy and minimizing communication overhead. However, susceptible model parameters render FL vulnerable to inference and impersonation attacks, necessitating strong and lightweight authentication protocols. This paper introduces a federated learning-based collaborative authentication protocol that incorporates anonymous mutual authentication and dynamic key agreement between vehicles and infrastructure nodes, enabling privacy-preserving data sharing while eliminating reliance on centralized entities. The protocol is evaluated using three aggregation strategies: FedAvg, FedProx, and FedOpt under varying client participation levels and epoch configurations. Experimental results demonstrate that FedOpt achieves superior convergence and generalization, reducing average loss by approximately 7.0% compared to FedAvg and 6.5% compared to FedProx, particularly in heterogeneous environments. Security analysis under the stochastic predictive machine model further validates robustness against adversarial threats. Overall, the proposed protocol enhances scalability, reduces propagation delay, and ensures efficient, privacy-preserving authentication in dynamic SIoV networks.
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (14) , 31809-31822
PQ-AuthV: Post-Quantum Secure Authentication With Aggregated Signatures in IoT-Enabled Smart Vehicle Networks
In the era of emerging vehicular technologies, Vehicular Ad-Hoc Networks (VANETs) play an important role in enhancing road safety as well as optimizing traffic management. With the rapid progress in Internet of Things (IoT), vehicles are increasingly interconnected, resulting in IoT-enabled smart vehicular networks that require secure communication mechanisms. Without secure VANET communication, real-time traffic information sharing can be limited, which compromises driver assistance systems and traffic flow efficiency. Moreover, existing classical cryptography-based schemes are also vulnerable to quantum attacks, while some post-quantum approaches overlook efficient verification and strong privacy protection as well. To address these limitations, this work proposes PQ-AuthV, a post-quantum secure authentication framework with aggregated signatures for IoT-driven vehicular communication systems. The proposed scheme enables mutual authentication, preserves conditional privacy, along with aggregate signature generation and validation. The security of the scheme is rigorously analyzed via the formal analysis in the Real-or-Random (ROR) oracle model, along with informal analysis, showing the robustness against various classical attacks, along with quantum attacks. Finally, a comprehensive evaluation shows that PQ-AuthV achieves robust security attributes with substantially lower computational and communication overheads for IoT-integrated vehicular networks compared to existing schemes.
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (14) , 31809-31822
PQ-AuthV: Post-Quantum Secure Authentication With Aggregated Signatures in IoT-Enabled Smart Vehicle Networks
In the era of emerging vehicular technologies, Vehicular Ad-Hoc Networks (VANETs) play an important role in enhancing road safety as well as optimizing traffic management. With the rapid progress in Internet of Things (IoT), vehicles are increasingly interconnected, resulting in IoT-enabled smart vehicular networks that require secure communication mechanisms. Without secure VANET communication, real-time traffic information sharing can be limited, which compromises driver assistance systems and traffic flow efficiency. Moreover, existing classical cryptography-based schemes are also vulnerable to quantum attacks, while some post-quantum approaches overlook efficient verification and strong privacy protection as well. To address these limitations, this work proposes PQ-AuthV, a post-quantum secure authentication framework with aggregated signatures for IoT-driven vehicular communication systems. The proposed scheme enables mutual authentication, preserves conditional privacy, along with aggregate signature generation and validation. The security of the scheme is rigorously analyzed via the formal analysis in the Real-or-Random (ROR) oracle model, along with informal analysis, showing the robustness against various classical attacks, along with quantum attacks. Finally, a comprehensive evaluation shows that PQ-AuthV achieves robust security attributes with substantially lower computational and communication overheads for IoT-integrated vehicular networks compared to existing schemes.
- Open Access
- 20 April 2026
- Published by Elsevier in SSRN Electronic Journal
Clinical Safety Validation of Interpretable Deep Learning for ESI-Free Emergency Department Triage Using a Dual-Attention Architecture
Background:AI-based triage models for emergency department (ED) overcrowding must ensure predictive performance, clinical safety, and interpretability. Prior studies suffer from target leakage by incorporating Emergency Severity Index (ESI) scores, or prioritize accuracy while failing to detect high-risk patients.Objective:To validate the clinical safety of an interpretable deep learning architecture (DA-V2) for ESI-free emergency department triage using outcome-based risk labels.Methods:Using MIMIC-IV-ED (N=407,735 adult visits), ESI scores were excluded from inputs and labels. Risk labels were defined by outcomes: High Risk (expired, transferred, or 30-day mortality), Medium Risk (admitted, survived >30 days), and Low Risk (discharged). DA-V2 integrates Squeeze-and-Excitation Feature Attention, Pre-LN Multi-Head Self-Attention, and Gated Residual Fusion, trained with focal loss and mixup augmentation in a five-seed ensemble. Seven baselines were evaluated under balanced and default class weighting, plus an MLP with DA-V2's identical pipeline as a training strategy control. Attention faithfulness was assessed via Spearman correlation with SHAP values.Results:DA-V2 achieved AUROC 0.7949 (95% CI: 0.7908-0.7991), significantly outperforming all balanced tree-based baselines (p<0.001). An MLP trained with the identical pipeline (MLP+ Enhanced) achieved statistically indistinguishable AUROC (0.7937, p=0.928) and equivalent High Risk (HR) Recall (74.2% vs. 74.1%), establishing focal loss, mixup, and seed ensembling as a generalizable training framework; DA-V2's unique advantage lies in verified interpretability. Default models-despite AUROC 0.796-0.800-detected fewer than 1% of high-risk patients. The dual verification protocol confirmed attention-SHAP alignment (Spearman rho=0.708, p < 10^-12). Ablation showed SE Feature Attention removal reduced HR Recall by 6.6 percentage points.Conclusions:DA-V2 demonstrates that clinically safe, interpretable AI for emergency triage is achievable without ESI scores. The training strategy drives clinical sensitivity, while DA-V2's attention mechanism uniquely enables verified interpretability, advancing deep learning as a trustworthy clinical decision support system.
- 20 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessClinical Safety Validation of Interpretable Deep Learning for ESI-Free Emergency Department Triage Using a Dual-Attention Architecture
Background:AI-based triage models for emergency department (ED) overcrowding must ensure predictive performance, clinical safety, and interpretability. Prior studies suffer from target leakage by incorporating Emergency Severity Index (ESI) scores, or prioritize accuracy while failing to detect high-risk patients.Objective:To validate the clinical safety of an interpretable deep learning architecture (DA-V2) for ESI-free emergency department triage using outcome-based risk labels.Methods:Using MIMIC-IV-ED (N=407,735 adult visits), ESI scores were excluded from inputs and labels. Risk labels were defined by outcomes: High Risk (expired, transferred, or 30-day mortality), Medium Risk (admitted, survived >30 days), and Low Risk (discharged). DA-V2 integrates Squeeze-and-Excitation Feature Attention, Pre-LN Multi-Head Self-Attention, and Gated Residual Fusion, trained with focal loss and mixup augmentation in a five-seed ensemble. Seven baselines were evaluated under balanced and default class weighting, plus an MLP with DA-V2's identical pipeline as a training strategy control. Attention faithfulness was assessed via Spearman correlation with SHAP values.Results:DA-V2 achieved AUROC 0.7949 (95% CI: 0.7908-0.7991), significantly outperforming all balanced tree-based baselines (p<0.001). An MLP trained with the identical pipeline (MLP+ Enhanced) achieved statistically indistinguishable AUROC (0.7937, p=0.928) and equivalent High Risk (HR) Recall (74.2% vs. 74.1%), establishing focal loss, mixup, and seed ensembling as a generalizable training framework; DA-V2's unique advantage lies in verified interpretability. Default models-despite AUROC 0.796-0.800-detected fewer than 1% of high-risk patients. The dual verification protocol confirmed attention-SHAP alignment (Spearman rho=0.708, p < 10^-12). Ablation showed SE Feature Attention removal reduced HR Recall by 6.6 percentage points.Conclusions:DA-V2 demonstrates that clinically safe, interpretable AI for emergency triage is achievable without ESI scores. The training strategy drives clinical sensitivity, while DA-V2's attention mechanism uniquely enables verified interpretability, advancing deep learning as a trustworthy clinical decision support system.
- Open Access
- 25 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (4) , 7831-7847
S-Auth: Schnorr-Enhanced Authentication Scheme for Security and Efficiency in Blockchain Web3.0
Web3 authentication stacks largely inherit ECDSA centric single-signature workflows that limit security and efficiency, while custody of identity data often remains application controlled rather than self-sovereign. We present S-Auth, an authentication layer that combines BIP340 Schnorr signatures with Decentralized Identifiers (DID), Verifiable Credentials (VC), and Content addressing (CID). The proposed solution utilizes Schnorr digital signatures, which have demonstrated improved security and efficiency over traditional schemes. The contributions of this work are as follows. Firstly, we apply the BIP340 standard to Schnorr digital signatures, bolstering security against various attacks including strong unforgeability under chosen message attack (SUF-CMA), non-malleability, linearity, related-key at tacks, hash collision, fault injection, nonce exfiltration, resource exhaustion, and domain separation. Secondly, we leverage the linearity property of Schnorr signatures to enable multi-signature aggregation and batch verification, addressing the inefficiency of existing schemes that rely on single signatures and thereby also enhancing privacy. Third, we combine the blockchain with DID, VC, and IPFS to provide a secure and self-sovereign identity that can be authenticated. Experiments comparing ECDSA, Ed25519, Schnorr, and BIP340 show that S-Auth reduces signature artifacts via aggregation, improves verifier throughput with batching, and decreases anchoring overhead while preserving user-controlled identity. S-Auth provides a self-sovereign, efficient, and secure authentication mechanism suitable for Web3 environments.
- 25 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 23 (4) , 7831-7847
Open AccessS-Auth: Schnorr-Enhanced Authentication Scheme for Security and Efficiency in Blockchain Web3.0
Web3 authentication stacks largely inherit ECDSA centric single-signature workflows that limit security and efficiency, while custody of identity data often remains application controlled rather than self-sovereign. We present S-Auth, an authentication layer that combines BIP340 Schnorr signatures with Decentralized Identifiers (DID), Verifiable Credentials (VC), and Content addressing (CID). The proposed solution utilizes Schnorr digital signatures, which have demonstrated improved security and efficiency over traditional schemes. The contributions of this work are as follows. Firstly, we apply the BIP340 standard to Schnorr digital signatures, bolstering security against various attacks including strong unforgeability under chosen message attack (SUF-CMA), non-malleability, linearity, related-key at tacks, hash collision, fault injection, nonce exfiltration, resource exhaustion, and domain separation. Secondly, we leverage the linearity property of Schnorr signatures to enable multi-signature aggregation and batch verification, addressing the inefficiency of existing schemes that rely on single signatures and thereby also enhancing privacy. Third, we combine the blockchain with DID, VC, and IPFS to provide a secure and self-sovereign identity that can be authenticated. Experiments comparing ECDSA, Ed25519, Schnorr, and BIP340 show that S-Auth reduces signature artifacts via aggregation, improves verifier throughput with batching, and decreases anchoring overhead while preserving user-controlled identity. S-Auth provides a self-sovereign, efficient, and secure authentication mechanism suitable for Web3 environments.
- Open Access
- 9 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (9) , 19141-19156
ZK-Sandbox: Zero-Knowledge Data Sandbox System for Enhancing End-to-End Security and Privacy of Verifiable Credentials
The rapid adoption of verifiable credentials (VCs) has intensified privacy and security challenges in digital verification, as traditional systems often require full credential disclosure, creating privacy risks and expanding the attack surface. Ensuring end-to-end privacy, security, and verifiability in such systems remains a significant challenge. This article introduces ZK-Sandbox, a zero-knowledge (ZK) Data Sandbox System that integrates ZK proofs (ZKPs), decentralized identifiers (DIDs), and blockchain anchoring to enable credential validation and verifiable badge issuance without exposing underlying data. ZK-Sandbox supports complex predicate evaluation by securely aggregating multiple VCs from trusted issuers, processing them via JSON Web Token signature verification, Circom-based zk-SNARK circuits, and Docker-isolated execution. The system issues VC-compatible ZK-Badges, cryptographically bound to a Poseidon hash and anchored to a blockchain-registered DID, containing only abstracted verification results. Experimental evaluation confirms 100% validation accuracy, complete detection of tampered submissions, and efficient performance–averaging 326-ms issuance latency, 570-ms off-chain verification, and 4.45-s on-chain verification. These results demonstrate that ZK-Sandbox is a privacy-by-design, scalable, and regulation-aligned solution for self-sovereign digital credential ecosystems.
- 9 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (9) , 19141-19156
Open AccessZK-Sandbox: Zero-Knowledge Data Sandbox System for Enhancing End-to-End Security and Privacy of Verifiable Credentials
The rapid adoption of verifiable credentials (VCs) has intensified privacy and security challenges in digital verification, as traditional systems often require full credential disclosure, creating privacy risks and expanding the attack surface. Ensuring end-to-end privacy, security, and verifiability in such systems remains a significant challenge. This article introduces ZK-Sandbox, a zero-knowledge (ZK) Data Sandbox System that integrates ZK proofs (ZKPs), decentralized identifiers (DIDs), and blockchain anchoring to enable credential validation and verifiable badge issuance without exposing underlying data. ZK-Sandbox supports complex predicate evaluation by securely aggregating multiple VCs from trusted issuers, processing them via JSON Web Token signature verification, Circom-based zk-SNARK circuits, and Docker-isolated execution. The system issues VC-compatible ZK-Badges, cryptographically bound to a Poseidon hash and anchored to a blockchain-registered DID, containing only abstracted verification results. Experimental evaluation confirms 100% validation accuracy, complete detection of tampered submissions, and efficient performance–averaging 326-ms issuance latency, 570-ms off-chain verification, and 4.45-s on-chain verification. These results demonstrate that ZK-Sandbox is a privacy-by-design, scalable, and regulation-aligned solution for self-sovereign digital credential ecosystems.
- Open Access
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 8699-8717
GNNRec: A Graph-Based Neural Network Model for Efficient and Robust Product Recommendation
Product recommendation systems are fundamental to modern e-commerce platforms, yet existing approaches face significant challenges with data sparsity, cold-start problems, and scalability limitations in large-scale applications. To address these challenges, this paper introduces GNNRec, a graph-based neural network model for efficient and robust product recommendation. It is a comprehensive graph-based recommendation framework that extends the LightGCN architecture with multi-relational user-item interactions, heterogeneous node embeddings, and multi-task learning capabilities. Our model incorporates rich semantic features and edge-weighted graph propagation to effectively capture complex user preferences and item characteristics. GNNRec addresses the cold-start problem through sophisticated graph propagation techniques and content-based initialization strategies, and handles data sparsity via adaptive sampling strategies. Extensive experiments on large-scale real-world datasets demonstrate that GNNRec achieves significant improvements over state-of-the-art baselines, achieving an AUC of 0.916 and consistent superior performance across all ranking metrics. Notably, GNNRec achieves NDCG@10 of 0.0428, representing a 4.4% improvement over the strongest baseline (NGCF at 0.0410). Moreover, GNNRec demonstrates strong robustness under data sparsity (99.988%), cold-start conditions (16.3% improvement for users with 5-10 interactions, and viable recommendations for zero-interaction users), and temporal dynamics, while achieving 42.4% faster training with 33.3% lower memory usage, confirming its scalability and effectiveness for practical deployment.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 8699-8717
Open AccessGNNRec: A Graph-Based Neural Network Model for Efficient and Robust Product Recommendation
Product recommendation systems are fundamental to modern e-commerce platforms, yet existing approaches face significant challenges with data sparsity, cold-start problems, and scalability limitations in large-scale applications. To address these challenges, this paper introduces GNNRec, a graph-based neural network model for efficient and robust product recommendation. It is a comprehensive graph-based recommendation framework that extends the LightGCN architecture with multi-relational user-item interactions, heterogeneous node embeddings, and multi-task learning capabilities. Our model incorporates rich semantic features and edge-weighted graph propagation to effectively capture complex user preferences and item characteristics. GNNRec addresses the cold-start problem through sophisticated graph propagation techniques and content-based initialization strategies, and handles data sparsity via adaptive sampling strategies. Extensive experiments on large-scale real-world datasets demonstrate that GNNRec achieves significant improvements over state-of-the-art baselines, achieving an AUC of 0.916 and consistent superior performance across all ranking metrics. Notably, GNNRec achieves NDCG@10 of 0.0428, representing a 4.4% improvement over the strongest baseline (NGCF at 0.0410). Moreover, GNNRec demonstrates strong robustness under data sparsity (99.988%), cold-start conditions (16.3% improvement for users with 5-10 interactions, and viable recommendations for zero-interaction users), and temporal dynamics, while achieving 42.4% faster training with 33.3% lower memory usage, confirming its scalability and effectiveness for practical deployment.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 171908-171927
UGC-Owner: Composable Copyright Authority Verification for AI-Generated Content Using NFT and Recursive Zero-Knowledge Proofs
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 171908-171927
Open AccessUGC-Owner: Composable Copyright Authority Verification for AI-Generated Content Using NFT and Recursive Zero-Knowledge Proofs