Kuang-Ching Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Kuang-Ching Wang has a total of 85 co-authors. Here are the top co-authors by number of shared publications:
- Caleb Linduff
Department of Electrical and Computer Engineering Clemson University 29634-0915 Clemson SC USA
3 shared publications - Ronald W. Gimbel
Clemson University
3 shared publications - Snigdhaswin Kar
ECE Department, Clemson University, Clemson, USA
3 shared publications - Christopher E. Gainey
Department of Emergency Medical Services Prisma Health Richland Hospital Columbia, SC United States
2 shared publications - D. Hudson Smith
Clemson University
2 shared publications - Gaurav Bagwe
Clemson University
2 shared publications
Latest Publications
- Open Access
Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks
Retrieval-augmented generation (RAG) enhances factual grounding by integrating retrieval mechanisms with generative models but introduces new attack surfaces, particularly through backdoor attacks. While prior research has largely focused on disinformation threats, fairness vulnerabilities remain underexplored. Unlike conventional backdoors that rely on direct trigger-to-target mappings, fairness-driven attacks exploit the interaction between retrieval and generation models, manipulating semantic relationships between target groups and social biases to establish a persistent and covert influence on content generation. This paper introduces BiasRAG, a systematic framework that exposes fairness vulnerabilities in RAG through a two-phase backdoor attack. During the pre-training phase, the query encoder is compromised to align the target group with the intended social bias, ensuring long-term persistence. In the post-deployment phase, adversarial documents are injected into knowledge bases to reinforce the backdoor, subtly influencing retrieved content while remaining undetectable under standard fairness evaluations. Together, BiasRAG ensures precise target alignment over sensitive attributes, stealthy execution, and resilience. Empirical evaluations demonstrate that BiasRAG achieves high attack success rates while preserving contextual relevance and utility, establishing a persistent and evolving threat to fairness in RAG.
Open AccessYour RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks
Retrieval-augmented generation (RAG) enhances factual grounding by integrating retrieval mechanisms with generative models but introduces new attack surfaces, particularly through backdoor attacks. While prior research has largely focused on disinformation threats, fairness vulnerabilities remain underexplored. Unlike conventional backdoors that rely on direct trigger-to-target mappings, fairness-driven attacks exploit the interaction between retrieval and generation models, manipulating semantic relationships between target groups and social biases to establish a persistent and covert influence on content generation. This paper introduces BiasRAG, a systematic framework that exposes fairness vulnerabilities in RAG through a two-phase backdoor attack. During the pre-training phase, the query encoder is compromised to align the target group with the intended social bias, ensuring long-term persistence. In the post-deployment phase, adversarial documents are injected into knowledge bases to reinforce the backdoor, subtly influencing retrieved content while remaining undetectable under standard fairness evaluations. Together, BiasRAG ensures precise target alignment over sensitive attributes, stealthy execution, and resilience. Empirical evaluations demonstrate that BiasRAG achieves high attack success rates while preserving contextual relevance and utility, establishing a persistent and evolving threat to fairness in RAG.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 15930-15948
Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks
Gaurav Bagwe, Saket Sanjeev Chaturvedi, Xiaolong Ma, Xiaoyong Yuan, Kuang-Ching Wang, Lan Emily Zhang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 15930-15948
Open AccessYour RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks
Gaurav Bagwe, Saket Sanjeev Chaturvedi, Xiaolong Ma, Xiaoyong Yuan, Kuang-Ching Wang, Lan Emily Zhang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- Open Access
Container Resource Allocation versus Performance of Data-intensive Applications on Different Cloud Servers
In recent years, data-intensive applications have been increasingly deployed on cloud systems. Such applications utilize significant compute, memory, and I/O resources to process large volumes of data. Optimizing the performance and cost-efficiency for such applications is a non-trivial problem. The problem becomes even more challenging with the increasing use of containers, which are popular due to their lower operational overheads and faster boot speed at the cost of weaker resource assurances for the hosted applications. In this paper, two containerized data-intensive applications with very different performance objectives and resource needs were studied on cloud servers with Docker containers running on Intel Xeon E5 and AMD EPYC Rome multi-core processors with a range of CPU, memory, and I/O configurations. Primary findings from our experiments include: 1) Allocating multiple cores to a compute-intensive application can improve performance, but only if the cores do not contend for the same caches, and the optimal core counts depend on the specific workload; 2) allocating more memory to a memory-intensive application than its deterministic data workload does not further improve performance; however, 3) having multiple such memory-intensive containers on the same server can lead to cache and memory bus contention leading to significant and volatile performance degradation. The comparative observations on Intel and AMD servers provided insights into trade-offs between larger numbers of distributed chiplets interconnected with higher speed buses (AMD) and larger numbers of centrally integrated cores and caches with lesser speed buses (Intel). For the two types of applications studied, the more distributed caches and faster data buses have benefited the deployment of larger numbers of containers.
Open AccessContainer Resource Allocation versus Performance of Data-intensive Applications on Different Cloud Servers
In recent years, data-intensive applications have been increasingly deployed on cloud systems. Such applications utilize significant compute, memory, and I/O resources to process large volumes of data. Optimizing the performance and cost-efficiency for such applications is a non-trivial problem. The problem becomes even more challenging with the increasing use of containers, which are popular due to their lower operational overheads and faster boot speed at the cost of weaker resource assurances for the hosted applications. In this paper, two containerized data-intensive applications with very different performance objectives and resource needs were studied on cloud servers with Docker containers running on Intel Xeon E5 and AMD EPYC Rome multi-core processors with a range of CPU, memory, and I/O configurations. Primary findings from our experiments include: 1) Allocating multiple cores to a compute-intensive application can improve performance, but only if the cores do not contend for the same caches, and the optimal core counts depend on the specific workload; 2) allocating more memory to a memory-intensive application than its deterministic data workload does not further improve performance; however, 3) having multiple such memory-intensive containers on the same server can lead to cache and memory bus contention leading to significant and volatile performance degradation. The comparative observations on Intel and AMD servers provided insights into trade-offs between larger numbers of distributed chiplets interconnected with higher speed buses (AMD) and larger numbers of centrally integrated cores and caches with lesser speed buses (Intel). For the two types of applications studied, the more distributed caches and faster data buses have benefited the deployment of larger numbers of containers.
- Open Access
- 12 October 2021
- Published by JMIR Publications Inc. in JMIR mHealth and uHealth
- Vol. 9  (10) , e32301
Complete and Resilient Documentation for Operational Medical Environments Leveraging Mobile Hands-free Technology in a Systems Approach: Experimental Study
Background: Prehospitalization documentation is a challenging task and prone to loss of information, as paramedics operate under disruptive environments requiring their constant attention to the patients. Objective: The aim of this study is to develop a mobile platform for hands-free prehospitalization documentation to assist first responders in operational medical environments by aggregating all existing solutions for noise resiliency and domain adaptation. Methods: The platform was built to extract meaningful medical information from the real-time audio streaming at the point of injury and transmit complete documentation to a field hospital prior to patient arrival. To this end, the state-of-the-art automatic speech recognition (ASR) solutions with the following modular improvements were thoroughly explored: noise-resilient ASR, multi-style training, customized lexicon, and speech enhancement. The development of the platform was strictly guided by qualitative research and simulation-based evaluation to address the relevant challenges through progressive improvements at every process step of the end-to-end solution. The primary performance metrics included medical word error rate (WER) in machine-transcribed text output and an F1 score calculated by comparing the autogenerated documentation to manual documentation by physicians. Results: The total number of 15,139 individual words necessary for completing the documentation were identified from all conversations that occurred during the physician-supervised simulation drills. The baseline model presented a suboptimal performance with a WER of 69.85% and an F1 score of 0.611. The noise-resilient ASR, multi-style training, and customized lexicon improved the overall performance; the finalized platform achieved a medical WER of 33.3% and an F1 score of 0.81 when compared to manual documentation. The speech enhancement degraded performance with medical WER increased from 33.3% to 46.33% and the corresponding F1 score decreased from 0.81 to 0.78. All changes in performance were statistically significant (P<.001). Conclusions: This study presented a fully functional mobile platform for hands-free prehospitalization documentation in operational medical environments and lessons learned from its implementation.
- 12 October 2021
- Published by JMIR Publications Inc. in JMIR mHealth and uHealth
- Vol. 9  (10) , e32301
Open AccessComplete and Resilient Documentation for Operational Medical Environments Leveraging Mobile Hands-free Technology in a Systems Approach: Experimental Study
Background: Prehospitalization documentation is a challenging task and prone to loss of information, as paramedics operate under disruptive environments requiring their constant attention to the patients. Objective: The aim of this study is to develop a mobile platform for hands-free prehospitalization documentation to assist first responders in operational medical environments by aggregating all existing solutions for noise resiliency and domain adaptation. Methods: The platform was built to extract meaningful medical information from the real-time audio streaming at the point of injury and transmit complete documentation to a field hospital prior to patient arrival. To this end, the state-of-the-art automatic speech recognition (ASR) solutions with the following modular improvements were thoroughly explored: noise-resilient ASR, multi-style training, customized lexicon, and speech enhancement. The development of the platform was strictly guided by qualitative research and simulation-based evaluation to address the relevant challenges through progressive improvements at every process step of the end-to-end solution. The primary performance metrics included medical word error rate (WER) in machine-transcribed text output and an F1 score calculated by comparing the autogenerated documentation to manual documentation by physicians. Results: The total number of 15,139 individual words necessary for completing the documentation were identified from all conversations that occurred during the physician-supervised simulation drills. The baseline model presented a suboptimal performance with a WER of 69.85% and an F1 score of 0.611. The noise-resilient ASR, multi-style training, and customized lexicon improved the overall performance; the finalized platform achieved a medical WER of 33.3% and an F1 score of 0.81 when compared to manual documentation. The speech enhancement degraded performance with medical WER increased from 33.3% to 46.33% and the corresponding F1 score decreased from 0.81 to 0.78. All changes in performance were statistically significant (P<.001). Conclusions: This study presented a fully functional mobile platform for hands-free prehospitalization documentation in operational medical environments and lessons learned from its implementation.
- Open Access
- 23 July 2021
- Published by JMIR Publications Inc. in JMIR Preprints
Complete and Resilient Documentation for Operational Medical Environments Leveraging Mobile Hands-free Technology in a Systems Approach: Experimental Study (Preprint)
BACKGROUND Prehospitalization documentation is a challenging task and prone to loss of information, as paramedics operate under disruptive environments requiring their constant attention to the patients. OBJECTIVE The aim of this study is to develop a mobile platform for hands-free prehospitalization documentation to assist first responders in operational medical environments by aggregating all existing solutions for noise resiliency and domain adaptation. METHODS The platform was built to extract meaningful medical information from the real-time audio streaming at the point of injury and transmit complete documentation to a field hospital prior to patient arrival. To this end, the state-of-the-art automatic speech recognition (ASR) solutions with the following modular improvements were thoroughly explored: noise-resilient ASR, multi-style training, customized lexicon, and speech enhancement. The development of the platform was strictly guided by qualitative research and simulation-based evaluation to address the relevant challenges through progressive improvements at every process step of the end-to-end solution. The primary performance metrics included medical word error rate (WER) in machine-transcribed text output and an F1 score calculated by comparing the autogenerated documentation to manual documentation by physicians. RESULTS The total number of 15,139 individual words necessary for completing the documentation were identified from all conversations that occurred during the physician-supervised simulation drills. The baseline model presented a suboptimal performance with a WER of 69.85% and an F1 score of 0.611. The noise-resilient ASR, multi-style training, and customized lexicon improved the overall performance; the finalized platform achieved a medical WER of 33.3% and an F1 score of 0.81 when compared to manual documentation. The speech enhancement degraded performance with medical WER increased from 33.3% to 46.33% and the corresponding F1 score decreased from 0.81 to 0.78. All changes in performance were statistically significant (P<.001). CONCLUSIONS This study presented a fully functional mobile platform for hands-free prehospitalization documentation in operational medical environments and lessons learned from its implementation.
- 23 July 2021
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessComplete and Resilient Documentation for Operational Medical Environments Leveraging Mobile Hands-free Technology in a Systems Approach: Experimental Study (Preprint)
BACKGROUND Prehospitalization documentation is a challenging task and prone to loss of information, as paramedics operate under disruptive environments requiring their constant attention to the patients. OBJECTIVE The aim of this study is to develop a mobile platform for hands-free prehospitalization documentation to assist first responders in operational medical environments by aggregating all existing solutions for noise resiliency and domain adaptation. METHODS The platform was built to extract meaningful medical information from the real-time audio streaming at the point of injury and transmit complete documentation to a field hospital prior to patient arrival. To this end, the state-of-the-art automatic speech recognition (ASR) solutions with the following modular improvements were thoroughly explored: noise-resilient ASR, multi-style training, customized lexicon, and speech enhancement. The development of the platform was strictly guided by qualitative research and simulation-based evaluation to address the relevant challenges through progressive improvements at every process step of the end-to-end solution. The primary performance metrics included medical word error rate (WER) in machine-transcribed text output and an F1 score calculated by comparing the autogenerated documentation to manual documentation by physicians. RESULTS The total number of 15,139 individual words necessary for completing the documentation were identified from all conversations that occurred during the physician-supervised simulation drills. The baseline model presented a suboptimal performance with a WER of 69.85% and an F1 score of 0.611. The noise-resilient ASR, multi-style training, and customized lexicon improved the overall performance; the finalized platform achieved a medical WER of 33.3% and an F1 score of 0.81 when compared to manual documentation. The speech enhancement degraded performance with medical WER increased from 33.3% to 46.33% and the corresponding F1 score decreased from 0.81 to 0.78. All changes in performance were statistically significant (P<.001). CONCLUSIONS This study presented a fully functional mobile platform for hands-free prehospitalization documentation in operational medical environments and lessons learned from its implementation.
- Open Access
- 1 May 2021
- Published by Western Journal of Emergency Medicine in Western Journal of Emergency Medicine
- Vol. 22  (3) , 636-643
Diabetes Patient Surveillance in the Emergency Department: Proof of Concept and Opportunities
Introduction: The purpose of this study was to characterize the at-risk diabetes and prediabetes patient population visiting emergency department (ED) and urgent care (UC) centers in upstate South Carolina. Methods: We conducted this retrospective study at the largest non-profit healthcare system in South Carolina, using electronic health record (EHR) data of patients who had an ED or UC visit between February 2, 2016-July 31, 2018. Key variables including International Classification of Diseases, 10th Revision codes, laboratory test results, family history, medication, and demographic characteristics were used to classify the patients as healthy, having prediabetes, having diabetes, being at-risk for prediabetes, or being at-risk for diabetes. Patients who were known to have diabetes were classified further as having controlled diabetes, management challenged, or uncontrolled diabetes. Population analysis was stratified by the patient's annual number of ED/UC visits. Results: The risk stratification revealed 4.58% unique patients with unrecognized diabetes and 10.34% of the known patients with diabetes considered to be suboptimaly controlled. Patients identified as diabetes management challenged had more ED/UC visits. Of note, 33.95% of the patients had unrecognized prediabetes/diabetes risk factors identified during their ED/UC with 87.95% having some form of healthcare insurance. Conclusion: This study supports the idea that a single ED/UC unscheduled visit can identify individuals with unrecognized diabetes and an at-risk prediabetes population using EHR data. A patient's ED/ UC visit, regardless of their primary reason for seeking care, may be an opportunity to provide early identification and diabetes disease management enrollment to augment the medical care of our community.
- 1 May 2021
- Published by Western Journal of Emergency Medicine in Western Journal of Emergency Medicine
- Vol. 22  (3) , 636-643
Open AccessDiabetes Patient Surveillance in the Emergency Department: Proof of Concept and Opportunities
Introduction: The purpose of this study was to characterize the at-risk diabetes and prediabetes patient population visiting emergency department (ED) and urgent care (UC) centers in upstate South Carolina. Methods: We conducted this retrospective study at the largest non-profit healthcare system in South Carolina, using electronic health record (EHR) data of patients who had an ED or UC visit between February 2, 2016-July 31, 2018. Key variables including International Classification of Diseases, 10th Revision codes, laboratory test results, family history, medication, and demographic characteristics were used to classify the patients as healthy, having prediabetes, having diabetes, being at-risk for prediabetes, or being at-risk for diabetes. Patients who were known to have diabetes were classified further as having controlled diabetes, management challenged, or uncontrolled diabetes. Population analysis was stratified by the patient's annual number of ED/UC visits. Results: The risk stratification revealed 4.58% unique patients with unrecognized diabetes and 10.34% of the known patients with diabetes considered to be suboptimaly controlled. Patients identified as diabetes management challenged had more ED/UC visits. Of note, 33.95% of the patients had unrecognized prediabetes/diabetes risk factors identified during their ED/UC with 87.95% having some form of healthcare insurance. Conclusion: This study supports the idea that a single ED/UC unscheduled visit can identify individuals with unrecognized diabetes and an at-risk prediabetes population using EHR data. A patient's ED/ UC visit, regardless of their primary reason for seeking care, may be an opportunity to provide early identification and diabetes disease management enrollment to augment the medical care of our community.
- Open Access
The Future of CISE Distributed Research Infrastructure
Shared research infrastructure that is globally distributed and widely accessible has been a hallmark of the networking community. This paper presents an initial snapshot of a vision for a possible future of mid-scale distributed research infrastructure aimed at enabling new types of research and discoveries. The paper is written from the perspective of "lessons learned" in constructing and operating the Global Environment for Network Innovations (GENI) infrastructure and attempts to project future concepts and solutions based on these lessons. The goal of this paper is to engage the community to contribute new ideas and to inform funding agencies about future research directions to realize this vision.
Open AccessThe Future of CISE Distributed Research Infrastructure
Shared research infrastructure that is globally distributed and widely accessible has been a hallmark of the networking community. This paper presents an initial snapshot of a vision for a possible future of mid-scale distributed research infrastructure aimed at enabling new types of research and discoveries. The paper is written from the perspective of "lessons learned" in constructing and operating the Global Environment for Network Innovations (GENI) infrastructure and attempts to project future concepts and solutions based on these lessons. The goal of this paper is to engage the community to contribute new ideas and to inform funding agencies about future research directions to realize this vision.
- Open Access
- 7 June 2017
- Published by Association for Computing Machinery (ACM)
- p. 129-131
Poster
Firewalls have been typically used to enforce network access control. Network Functions Virtualization (NFV) envisions to implement firewall function as software instance (a.k.a virtual firewall). Virtual firewall provides great flexibility and elasticity, which are necessary to protect virtualized environments. In this poster, we propose an innovative virtual firewall controller, VFW Controller, which enables safe, efficient and cost-effective virtual firewall elasticity control. In addition, we implement the core components of VFW Controller on top of NFV and SDN environments. Our experimental results demonstrate that VFW Controller is efficient to provide safe elasticity control of virtual firewalls.
- 7 June 2017
- Published by Association for Computing Machinery (ACM)
- p. 129-131
Open AccessPoster
Firewalls have been typically used to enforce network access control. Network Functions Virtualization (NFV) envisions to implement firewall function as software instance (a.k.a virtual firewall). Virtual firewall provides great flexibility and elasticity, which are necessary to protect virtualized environments. In this poster, we propose an innovative virtual firewall controller, VFW Controller, which enables safe, efficient and cost-effective virtual firewall elasticity control. In addition, we implement the core components of VFW Controller on top of NFV and SDN environments. Our experimental results demonstrate that VFW Controller is efficient to provide safe elasticity control of virtual firewalls.
- 1 May 2012
- Published by American Society of Civil Engineers (ASCE) in Journal of Transportation Engineering
- Vol. 138  (5) , 512-519
Online Traffic Surveillance: Impact of Wireless Communication on Video Quality
The current wireless technological advancements in video surveillance are bound to shape the future of online traffic monitoring applications for intelligent transportation systems. The authors present the effect of using 802.11 b/g wireless technology on real-time traffic surveillance systems. Although numerous studies have focused on studying the quality requirements for video streaming using a simulation environment, there is limited information on such performance in a real world environment. This study analyzed key video quality parameters, including packet jitter and packet rate, suggesting necessary guidelines for the transportation agency personnel to use when evaluating and selecting surveillance and communication tools. Study of the packet jitter showed that jitter and skipped video are strongly correlated with one another. Moreover, tolerated jitter ranges were identified that would ensure smooth traffic surveillance video. Findings indicated that to avoid jitter-related problems and ...
- 1 May 2012
- Published by American Society of Civil Engineers (ASCE) in Journal of Transportation Engineering
- Vol. 138  (5) , 512-519
Online Traffic Surveillance: Impact of Wireless Communication on Video Quality
The current wireless technological advancements in video surveillance are bound to shape the future of online traffic monitoring applications for intelligent transportation systems. The authors present the effect of using 802.11 b/g wireless technology on real-time traffic surveillance systems. Although numerous studies have focused on studying the quality requirements for video streaming using a simulation environment, there is limited information on such performance in a real world environment. This study analyzed key video quality parameters, including packet jitter and packet rate, suggesting necessary guidelines for the transportation agency personnel to use when evaluating and selecting surveillance and communication tools. Study of the packet jitter showed that jitter and skipped video are strongly correlated with one another. Moreover, tolerated jitter ranges were identified that would ensure smooth traffic surveillance video. Findings indicated that to avoid jitter-related problems and ...
- 10 August 2009
- Published by Taylor & Francis in Journal of Intelligent Transportation Systems
- Vol. 13  (3) , 111-126
A Framework for Performance Evaluation of Communication Alternatives for Intelligent Transportation Systems
The communications infrastructure is critical for effective and efficient operation of intelligent transportation system (ITS) functions. Relatively large investments from the public sector are usually required to construct and maintain or provide leasing fee to private telecommunication companies for such infrastructure. Consequently, a rigorous analysis and evaluation of communication alternatives is of utmost importance. This research sought to develop a systematic framework and associated platform that will facilitate the evaluation of the communication alternatives for ITS. The framework includes identifying alternative architectures, such as distributed, centralized, or a combination of both, and identifying communication mediums including wired, wireless, or a combination of both. This article presents, as a part of the framework, important measures of effectiveness (MOE) for making objective comparisons between alternatives based on the performance of the communication systems related to ITS. A simulation platform that integrated a traffic and communication simulator was used for efficiently evaluating communication alternatives with different architecture and mediums by generating important MOEs. A case study was performed for a test network in Greenville, South Carolina. The proposed evaluation framework identified four communication alternatives, namely, the centralized wired, distributed wired, centralized wireless, and distributed wireless, to generate the selected MOEs, such as throughput, delivery ratio, and throughput cost ratios, for comparing and analyzing these alternatives.
- 10 August 2009
- Published by Taylor & Francis in Journal of Intelligent Transportation Systems
- Vol. 13  (3) , 111-126
A Framework for Performance Evaluation of Communication Alternatives for Intelligent Transportation Systems
The communications infrastructure is critical for effective and efficient operation of intelligent transportation system (ITS) functions. Relatively large investments from the public sector are usually required to construct and maintain or provide leasing fee to private telecommunication companies for such infrastructure. Consequently, a rigorous analysis and evaluation of communication alternatives is of utmost importance. This research sought to develop a systematic framework and associated platform that will facilitate the evaluation of the communication alternatives for ITS. The framework includes identifying alternative architectures, such as distributed, centralized, or a combination of both, and identifying communication mediums including wired, wireless, or a combination of both. This article presents, as a part of the framework, important measures of effectiveness (MOE) for making objective comparisons between alternatives based on the performance of the communication systems related to ITS. A simulation platform that integrated a traffic and communication simulator was used for efficiently evaluating communication alternatives with different architecture and mediums by generating important MOEs. A case study was performed for a test network in Greenville, South Carolina. The proposed evaluation framework identified four communication alternatives, namely, the centralized wired, distributed wired, centralized wireless, and distributed wireless, to generate the selected MOEs, such as throughput, delivery ratio, and throughput cost ratios, for comparing and analyzing these alternatives.