Skip to main content

Junaid Qadir

Qatar University, Department Of Computer Science & Engineering, Faculty Member

Followers

38

Following

5

Public Views

Junaid Qadir completed Ph.D. from University of New South Wales, Australia in 2008 and his Bachelors in Electrical Engineering from UET, Lahore, Pakistan in 2000. He is an Associate Professor at the Information Technology University (ITU)—Punjab, Lahore since December 2015.

less

Interests

Uploads

Papers by Junaid Qadir

Advancing Education Through Extended Reality and Internet of Everything Enabled Metaverses: Applications, Challenges, and Open Issues

IEEE transactions on learning technologies, 2024

Motion Comfort Optimization for Autonomous Vehicles: Concepts, Methods, and Techniques

arXiv (Cornell University), Jun 15, 2023

Senior Member, IEEE casted to reach 186.40 billion USD by 2030, and in 2021 alone, the AV market ... more

Security and privacy of internet of medical things: A contemporary review in the age of surveillance, botnets, and adversarial ML

Journal of Network and Computer Applications, May 1, 2022

Machine learning based optimized live virtual machine migration over WAN links

Telecommunication Systems, May 25, 2016

Transfer Learning for Improving Speech Emotion Classification Accuracy

arXiv (Cornell University), Jan 19, 2018

The majority of existing speech emotion recognition research focuses on automatic emotion detecti... more The majority of existing speech emotion recognition research focuses on automatic emotion detection using training and testing data from same corpus collected under the same conditions. The performance of such systems has been shown to drop significantly in cross-corpus and cross-language scenarios. To address the problem, this paper exploits a transfer learning technique to improve the performance of speech emotion recognition systems that is novel in cross-language and cross-corpus scenarios. Evaluations on five different corpora in three different languages show that Deep Belief Networks (DBNs) offer better accuracy than previous approaches on cross-corpus emotion recognition, relative to a Sparse Autoencoder and SVM baseline system. Results also suggest that using a large number of languages for training and using a small fraction of the target data in training can significantly boost accuracy compared with baseline also for the corpus with limited training examples.

Cross-Language Speech Emotion Recognition Using Multimodal Dual Attention Transformers

arXiv (Cornell University), Jun 23, 2023

Outcome-Based Engineering Education: A Global Report of International OBE Accreditation and Assessment Practices

, where he directs the ICTD; Human Development; Systems; Big Data Analytics; Networks (IHSAN) Res... more , where he directs the ICTD; Human Development; Systems; Big Data Analytics; Networks (IHSAN) Research Lab. His primary research interests are in the areas of computer systems and networking, applied machine learning, using ICT for development (ICT4D); and engineering education. He is the author of more than 100 peer-reviewed research papers that have been published at various top conferences and journals. He currently serves as the Chairperson of the Electrical Engineering department at ITU. He was awarded the HEC Best University Teacher Award, the highest national teaching award in Pakistan, in 2012. He is a senior member of IEEE. He is an ACM Distinguished Speaker for a three-year term starting 2020 and an ACM Senior Member.

Membership Inference Attacks on DNNs using Adversarial Perturbations

arXiv (Cornell University), Jul 11, 2023

ChatGPT and Large Language Models (LLMs) in Healthcare: Opportunities and Risks

ChatGPT, a pre-trained large language model (LLM), has the potential to transform healthcare by p... more ChatGPT, a pre-trained large language model (LLM), has the potential to transform healthcare by providing valid clinical insights and reducing doctors' workload. There are already signs that such tools can be useful for automating the generation of patient discharge reports, clinical vignettes, and radiology reports. Such tools can also capture the vast medical knowledge base as demonstrated by ChatGPT clearing the United States Medical Licensing Examination (USMLE). Such tools promise to make healthcare more accessible, scalable, and efficient, leading to better patient outcomes. However, such tools are far from perfect and well-known to be susceptible to error, misinformation, and bias. In this paper, we review the potential applications of ChatGPT in healthcare and also identify potentials risks that must be addressed before ChatGPT and other LLM tools can be safely adopted in healthcare.

Consistent Valid Physically-Realizable Adversarial Attack against Crowd-flow Prediction Models

arXiv (Cornell University), Mar 5, 2023

Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow... more Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow patterns, which can be used for more effective urban planning and smart city management. However, DL models have been known to perform poorly on inconspicuous adversarial perturbations. Although many works have studied these adversarial perturbations in general, the adversarial vulnerabilities of deep crowd-flow prediction models in particular have remained largely unexplored. In this paper, we perform a rigorous analysis of the adversarial vulnerabilities of DL-based crowd-flow prediction models under multiple threat settings, making threefold contributions. (1) We propose CaV-detect by formally identifying two novel properties-Consistency and Validity-of the crowd-flow prediction inputs that enable the detection of standard adversarial inputs with 0% false acceptance rate (FAR). (2) We leverage universal adversarial perturbations and an adaptive adversarial loss to present adaptive adversarial attacks to evade CaV-detect defense. (3) We propose CVPR, a Consistent, Valid and Physically-Realizable adversarial attack, that explicitly inducts the consistency and validity priors in the perturbation generation mechanism. We find out that although the crowd-flow models are vulnerable to adversarial perturbations, it is extremely challenging to simulate these perturbations in physical settings, notably when CaVdetect is in place. We also show that CVPR attack considerably outperforms the adaptively modified standard attacks in FAR and adversarial loss metrics. We conclude with useful insights emerging from our work and highlight promising future research directions.

Soft Computing Techniques for Dependable Cyber-Physical Systems

arXiv (Cornell University), Jan 25, 2018

Challenges and Countermeasures for Adversarial Attacks on Deep Reinforcement Learning

arXiv (Cornell University), Jan 27, 2020

A Generative Model to Synthesize EEG Data for Epileptic Seizure Prediction

arXiv (Cornell University), Dec 1, 2020

Epilepsy, a group of neurological disorders, is characterized by recurrent and unpredictable seiz... more Epilepsy, a group of neurological disorders, is characterized by recurrent and unpredictable seizures, which can cause an unexpected loss of awareness for patients resulting in a risk of life-threatening incidents. Prediction of seizure before they occur is therefore vital for bringing normalcy to the lives of patients. For decades researchers have been employing machine learning (ML) techniques based on hand-crafted features for seizure prediction. However, ML methods are too complicated to select the best ML model or the best features. Deep Learning (DL) methods are beneficial in the sense of automatic feature extraction. Lamentably despite decades of research, epilepsy prediction remains an unresolved problem. One of the roadblocks for accurate seizure prediction, among others, is the scarcity of good quality epileptic seizure data. This paper addresses this problem by proposing a deep convolutional generative adversarial network (DCGAN) to generate synthetic EEG data samples. We use two methods to validate the synthesized data namely, one-class SVM and a new proposal which we refer to as convolutional epileptic seizure predictor (CESP). Another objective of our study is to evaluate the performance of wellknown deep learning models (e.g., VGG16, VGG19, ResNet50, and Inceptionv3) by training the models on augmented EEG data using transfer learning with average time of 10 min between true prediction and seizure onset. Our results show that the CESP model achieves sensitivity of 78.11% and 88.21%, and false prediction rate of 0.27/h and 0.14/h for the training on synthesized and testing on real Epilepsyecosystem and CHB-MIT datasets, respectively. The impressive performance of CESP trained on synthesized data shows that the synthetic data acquired the correlation between features and labels very well. We also show that using the employment of the idea of transfer learning and data augmentation in patient-specific manner provides the highest accuracy with sensitivity of 90.03% and 0.03 FPR/h which was achieved using Inceptionv3, and that augmenting the data with the samples generated from DCGAN increased the prediction results of our CESP model and Inceptionv3 by 4-5% as compared to the state-of-the-art traditional augmentation techniques. Finally, we note that the prediction results of CESP achieved by using the augmented data are better than the chance level for both datasets.

Single-Shot Retinal Image Enhancement Using Untrained and Pretrained Neural Networks Priors Integrated with Analytical Image Priors

Retinal images acquired using fundus cameras are often visually blurred due to imperfect imaging ... more Retinal images acquired using fundus cameras are often visually blurred due to imperfect imaging conditions, refractive medium turbidity, and motion blur. In addition, ocular diseases such as the presence of cataract also result in blurred retinal images. The presence of blur in retinal fundus images reduces the effectiveness of the diagnosis process of an expert ophthalmologist or a computer-aided detection/diagnosis system. In this paper, we put forward a single-shot deep image prior (DIP)-based approach for retinal image enhancement. Unlike typical deep learning-based approaches, our method does not require any training data. Instead, our DIP-based method can learn the underlying image prior while using a single degraded image. To perform retinal image enhancement, we frame it as a layer decomposition problem and investigate the use of two well-known analytical priors, i.e., dark channel prior (DCP) and bright channel prior (BCP) for atmospheric light estimation. We show that both the untrained neural networks and the pretrained neural networks can be used to generate an enhanced image while using only a single degraded image. We evaluate our proposed framework quantitatively on five datasets using three widely used metrics and complement that with a subjective qualitative assessment of the enhancement by two expert ophthalmologists. We have compared our method with a recent state-of-the-art method cofe-Net using synthetically degraded retinal fundus images and show that our method outperforms the state-ofthe-art method and provides a gain of 1.23 and 1.4 in average PSNR and SSIM respectively. Our method also outperforms other works proposed in the literature, which have evaluated their performance on non-public proprietary datasets, on the basis of the reported results.

Use of Transfer Learning and Wavelet Transform for Breast Cancer Detection

Analyzing Misinformation Through The Lens of Systems Thinking

Performance analysis of 802.11 DCF with limited channels

Wireless communication has achieved great proliferation in communication technologies. IEEE 802.1... more Wireless communication has achieved great proliferation in communication technologies. IEEE 802.11 WLANs have been widely deployed as wireless access mechanism technology. To keep up with growing bandwidth and data rate requirements, legacy standards such as IEEE 802.11 have been deeply researched for their operation in multi-channel environment through usage of multiple interfaces. In the multi-channel multi-interface environment, it is common to assume that number of available channels are greater than or equal to number of interfaces. We evaluate performance of IEEE 802.11 DCF (Distributed Coordination Function) in multi-channel multi-interface environment in which number of channels are less than number of interfaces; contradicting common assumption where number of channels are greater than or equal to number of interfaces. In our study, we present scenarios under which number of available channels can become less than total number of interfaces on a node to motivate the problem. Through evaluation of accurate analytical models, we suggest that it is not possible to achieve higher average network throughput by tuning multiple interfaces on one channel. By keeping aggregate traffic constant for a channel, greater number of interfaces on a channel increases collisions, resulting in less effective usage of channel.

Making federated learning robust to adversarial attacks by learning data and model association

Computers & Security, Oct 1, 2022

A Student Primer On How to Thrive in Post-COVID-19 Engineering Education

IEEE Access Special Section Editorial: Health Informatics for the Developing World

IEEE Access, 2017

Advancing Education Through Extended Reality and Internet of Everything Enabled Metaverses: Applications, Challenges, and Open Issues

IEEE transactions on learning technologies, 2024

Motion Comfort Optimization for Autonomous Vehicles: Concepts, Methods, and Techniques

arXiv (Cornell University), Jun 15, 2023

Senior Member, IEEE casted to reach 186.40 billion USD by 2030, and in 2021 alone, the AV market ... more

Security and privacy of internet of medical things: A contemporary review in the age of surveillance, botnets, and adversarial ML

Journal of Network and Computer Applications, May 1, 2022

Machine learning based optimized live virtual machine migration over WAN links

Telecommunication Systems, May 25, 2016

Transfer Learning for Improving Speech Emotion Classification Accuracy

arXiv (Cornell University), Jan 19, 2018

The majority of existing speech emotion recognition research focuses on automatic emotion detecti... more The majority of existing speech emotion recognition research focuses on automatic emotion detection using training and testing data from same corpus collected under the same conditions. The performance of such systems has been shown to drop significantly in cross-corpus and cross-language scenarios. To address the problem, this paper exploits a transfer learning technique to improve the performance of speech emotion recognition systems that is novel in cross-language and cross-corpus scenarios. Evaluations on five different corpora in three different languages show that Deep Belief Networks (DBNs) offer better accuracy than previous approaches on cross-corpus emotion recognition, relative to a Sparse Autoencoder and SVM baseline system. Results also suggest that using a large number of languages for training and using a small fraction of the target data in training can significantly boost accuracy compared with baseline also for the corpus with limited training examples.

Cross-Language Speech Emotion Recognition Using Multimodal Dual Attention Transformers

arXiv (Cornell University), Jun 23, 2023

Outcome-Based Engineering Education: A Global Report of International OBE Accreditation and Assessment Practices

, where he directs the ICTD; Human Development; Systems; Big Data Analytics; Networks (IHSAN) Res... more , where he directs the ICTD; Human Development; Systems; Big Data Analytics; Networks (IHSAN) Research Lab. His primary research interests are in the areas of computer systems and networking, applied machine learning, using ICT for development (ICT4D); and engineering education. He is the author of more than 100 peer-reviewed research papers that have been published at various top conferences and journals. He currently serves as the Chairperson of the Electrical Engineering department at ITU. He was awarded the HEC Best University Teacher Award, the highest national teaching award in Pakistan, in 2012. He is a senior member of IEEE. He is an ACM Distinguished Speaker for a three-year term starting 2020 and an ACM Senior Member.

Membership Inference Attacks on DNNs using Adversarial Perturbations

arXiv (Cornell University), Jul 11, 2023

ChatGPT and Large Language Models (LLMs) in Healthcare: Opportunities and Risks

ChatGPT, a pre-trained large language model (LLM), has the potential to transform healthcare by p... more ChatGPT, a pre-trained large language model (LLM), has the potential to transform healthcare by providing valid clinical insights and reducing doctors' workload. There are already signs that such tools can be useful for automating the generation of patient discharge reports, clinical vignettes, and radiology reports. Such tools can also capture the vast medical knowledge base as demonstrated by ChatGPT clearing the United States Medical Licensing Examination (USMLE). Such tools promise to make healthcare more accessible, scalable, and efficient, leading to better patient outcomes. However, such tools are far from perfect and well-known to be susceptible to error, misinformation, and bias. In this paper, we review the potential applications of ChatGPT in healthcare and also identify potentials risks that must be addressed before ChatGPT and other LLM tools can be safely adopted in healthcare.

Consistent Valid Physically-Realizable Adversarial Attack against Crowd-flow Prediction Models

arXiv (Cornell University), Mar 5, 2023

Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow... more Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow patterns, which can be used for more effective urban planning and smart city management. However, DL models have been known to perform poorly on inconspicuous adversarial perturbations. Although many works have studied these adversarial perturbations in general, the adversarial vulnerabilities of deep crowd-flow prediction models in particular have remained largely unexplored. In this paper, we perform a rigorous analysis of the adversarial vulnerabilities of DL-based crowd-flow prediction models under multiple threat settings, making threefold contributions. (1) We propose CaV-detect by formally identifying two novel properties-Consistency and Validity-of the crowd-flow prediction inputs that enable the detection of standard adversarial inputs with 0% false acceptance rate (FAR). (2) We leverage universal adversarial perturbations and an adaptive adversarial loss to present adaptive adversarial attacks to evade CaV-detect defense. (3) We propose CVPR, a Consistent, Valid and Physically-Realizable adversarial attack, that explicitly inducts the consistency and validity priors in the perturbation generation mechanism. We find out that although the crowd-flow models are vulnerable to adversarial perturbations, it is extremely challenging to simulate these perturbations in physical settings, notably when CaVdetect is in place. We also show that CVPR attack considerably outperforms the adaptively modified standard attacks in FAR and adversarial loss metrics. We conclude with useful insights emerging from our work and highlight promising future research directions.

Soft Computing Techniques for Dependable Cyber-Physical Systems

arXiv (Cornell University), Jan 25, 2018

Challenges and Countermeasures for Adversarial Attacks on Deep Reinforcement Learning

arXiv (Cornell University), Jan 27, 2020

A Generative Model to Synthesize EEG Data for Epileptic Seizure Prediction

arXiv (Cornell University), Dec 1, 2020

Epilepsy, a group of neurological disorders, is characterized by recurrent and unpredictable seiz... more Epilepsy, a group of neurological disorders, is characterized by recurrent and unpredictable seizures, which can cause an unexpected loss of awareness for patients resulting in a risk of life-threatening incidents. Prediction of seizure before they occur is therefore vital for bringing normalcy to the lives of patients. For decades researchers have been employing machine learning (ML) techniques based on hand-crafted features for seizure prediction. However, ML methods are too complicated to select the best ML model or the best features. Deep Learning (DL) methods are beneficial in the sense of automatic feature extraction. Lamentably despite decades of research, epilepsy prediction remains an unresolved problem. One of the roadblocks for accurate seizure prediction, among others, is the scarcity of good quality epileptic seizure data. This paper addresses this problem by proposing a deep convolutional generative adversarial network (DCGAN) to generate synthetic EEG data samples. We use two methods to validate the synthesized data namely, one-class SVM and a new proposal which we refer to as convolutional epileptic seizure predictor (CESP). Another objective of our study is to evaluate the performance of wellknown deep learning models (e.g., VGG16, VGG19, ResNet50, and Inceptionv3) by training the models on augmented EEG data using transfer learning with average time of 10 min between true prediction and seizure onset. Our results show that the CESP model achieves sensitivity of 78.11% and 88.21%, and false prediction rate of 0.27/h and 0.14/h for the training on synthesized and testing on real Epilepsyecosystem and CHB-MIT datasets, respectively. The impressive performance of CESP trained on synthesized data shows that the synthetic data acquired the correlation between features and labels very well. We also show that using the employment of the idea of transfer learning and data augmentation in patient-specific manner provides the highest accuracy with sensitivity of 90.03% and 0.03 FPR/h which was achieved using Inceptionv3, and that augmenting the data with the samples generated from DCGAN increased the prediction results of our CESP model and Inceptionv3 by 4-5% as compared to the state-of-the-art traditional augmentation techniques. Finally, we note that the prediction results of CESP achieved by using the augmented data are better than the chance level for both datasets.

Single-Shot Retinal Image Enhancement Using Untrained and Pretrained Neural Networks Priors Integrated with Analytical Image Priors

Retinal images acquired using fundus cameras are often visually blurred due to imperfect imaging ... more Retinal images acquired using fundus cameras are often visually blurred due to imperfect imaging conditions, refractive medium turbidity, and motion blur. In addition, ocular diseases such as the presence of cataract also result in blurred retinal images. The presence of blur in retinal fundus images reduces the effectiveness of the diagnosis process of an expert ophthalmologist or a computer-aided detection/diagnosis system. In this paper, we put forward a single-shot deep image prior (DIP)-based approach for retinal image enhancement. Unlike typical deep learning-based approaches, our method does not require any training data. Instead, our DIP-based method can learn the underlying image prior while using a single degraded image. To perform retinal image enhancement, we frame it as a layer decomposition problem and investigate the use of two well-known analytical priors, i.e., dark channel prior (DCP) and bright channel prior (BCP) for atmospheric light estimation. We show that both the untrained neural networks and the pretrained neural networks can be used to generate an enhanced image while using only a single degraded image. We evaluate our proposed framework quantitatively on five datasets using three widely used metrics and complement that with a subjective qualitative assessment of the enhancement by two expert ophthalmologists. We have compared our method with a recent state-of-the-art method cofe-Net using synthetically degraded retinal fundus images and show that our method outperforms the state-ofthe-art method and provides a gain of 1.23 and 1.4 in average PSNR and SSIM respectively. Our method also outperforms other works proposed in the literature, which have evaluated their performance on non-public proprietary datasets, on the basis of the reported results.

Use of Transfer Learning and Wavelet Transform for Breast Cancer Detection

Analyzing Misinformation Through The Lens of Systems Thinking

Performance analysis of 802.11 DCF with limited channels

Wireless communication has achieved great proliferation in communication technologies. IEEE 802.1... more Wireless communication has achieved great proliferation in communication technologies. IEEE 802.11 WLANs have been widely deployed as wireless access mechanism technology. To keep up with growing bandwidth and data rate requirements, legacy standards such as IEEE 802.11 have been deeply researched for their operation in multi-channel environment through usage of multiple interfaces. In the multi-channel multi-interface environment, it is common to assume that number of available channels are greater than or equal to number of interfaces. We evaluate performance of IEEE 802.11 DCF (Distributed Coordination Function) in multi-channel multi-interface environment in which number of channels are less than number of interfaces; contradicting common assumption where number of channels are greater than or equal to number of interfaces. In our study, we present scenarios under which number of available channels can become less than total number of interfaces on a node to motivate the problem. Through evaluation of accurate analytical models, we suggest that it is not possible to achieve higher average network throughput by tuning multiple interfaces on one channel. By keeping aggregate traffic constant for a channel, greater number of interfaces on a channel increases collisions, resulting in less effective usage of channel.

Making federated learning robust to adversarial attacks by learning data and model association

Computers & Security, Oct 1, 2022

A Student Primer On How to Thrive in Post-COVID-19 Engineering Education

IEEE Access Special Section Editorial: Health Informatics for the Developing World

IEEE Access, 2017