Rashmi Sahay
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Rashmi Sahay has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
Barsha Mitra
Department of Computer Science and Information Systems, BITS Pilani Hyderabad Campus, Hyderabad, India
8 shared publicationsG. Geethakumari
BITS-Pilani Hyderabad Campus
5 shared publications- G. Geethakumari
Department of CSIS, BITS Pilani, Hyderabad Campus, Hyderabad, India
5 shared publications - Kaushik Sekaran
ICFAI Foundation for Higher Education
2 shared publications Koushik Modugu
Department of Computer Science and Information Systems, Hyderabad Campus, BITS-Pilani
2 shared publications- Amina Salhi
Department of Information Systems, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University, Riyadh, Saudi Arabia
1 shared publication
Latest Publications
- 13 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (2) , 4932-4946
Trust-Centric Mitigation of Time-Warping Adversarial Attacks in Wearable Sensor Platforms
Wearable sensors are increasingly deployed in healthcare, fitness, and human activity recognition (HAR). Adversarial attacks on deep learning models used in HRA increase security concerns in consumer goods. Adversarial time-series perturbations decrease the accuracy of prediction models. This paper investigates the impact of Time-Warping Adversarial Attacks (TWAA), which manipulate the temporal axis of sensor signals while preserving amplitude ranges. Thereby, generating adversarially misleading samples. We propose a trust-centric framework comprising attack execution, detection, and mitigation modules, and evaluate it on two benchmark datasets (UCI HAR, WISDM and MHEALTH) using baseline deep models (CNN, LSTM, and CNN–LSTM). Experimental results show that TWAA significantly degrades classification performance. It reduces the adversarial accuracy by up to 30% and achieving attack success rates exceeding 40% at moderate perturbation parameters. Our detection mechanism achieves AUROC > 0.85 across different types of TWAA, while the mitigation strategy restores model accuracy to above 80% with inference latency under 0.01 ms/sample. These findings demonstrate both the severity of temporal adversarial threats in HAR and the effectiveness of our proposed trust-centric defense. Unlike existing mechanisms that rely on adversarial training or offline robustness enhancement, the proposed framework enables runtime detection of physically plausible time-warping adversarial attacks and applies conditional input-level temporal repair using dynamic time warping. Thus, providing a holistic pipeline for the analysis, detection and mitigation of TWAA on wearable consumer goods.
- 13 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (2) , 4932-4946
Trust-Centric Mitigation of Time-Warping Adversarial Attacks in Wearable Sensor Platforms
Wearable sensors are increasingly deployed in healthcare, fitness, and human activity recognition (HAR). Adversarial attacks on deep learning models used in HRA increase security concerns in consumer goods. Adversarial time-series perturbations decrease the accuracy of prediction models. This paper investigates the impact of Time-Warping Adversarial Attacks (TWAA), which manipulate the temporal axis of sensor signals while preserving amplitude ranges. Thereby, generating adversarially misleading samples. We propose a trust-centric framework comprising attack execution, detection, and mitigation modules, and evaluate it on two benchmark datasets (UCI HAR, WISDM and MHEALTH) using baseline deep models (CNN, LSTM, and CNN–LSTM). Experimental results show that TWAA significantly degrades classification performance. It reduces the adversarial accuracy by up to 30% and achieving attack success rates exceeding 40% at moderate perturbation parameters. Our detection mechanism achieves AUROC > 0.85 across different types of TWAA, while the mitigation strategy restores model accuracy to above 80% with inference latency under 0.01 ms/sample. These findings demonstrate both the severity of temporal adversarial threats in HAR and the effectiveness of our proposed trust-centric defense. Unlike existing mechanisms that rely on adversarial training or offline robustness enhancement, the proposed framework enables runtime detection of physically plausible time-warping adversarial attacks and applies conditional input-level temporal repair using dynamic time warping. Thus, providing a holistic pipeline for the analysis, detection and mitigation of TWAA on wearable consumer goods.
- 28 November 2024
- Published by Wiley in Internet Technology Letters
- Vol. 8 (5)
LiteEncrypt : A Lightweight Block Cipher for Secure Communication in IoT Enabled Sensor
The resource‐constrained characteristics of the Internet of Things (IoT) sensors lead to a pressing security requirement for developing lightweight cryptographic algorithms. This work introduces and analyzes a novel lightweight cryptographic algorithm for resource‐constrained devices. The algorithm is based on the Feistel structure in which the round function combines circular shifts, XOR, and key addition operation for data encryption and decryption. With a key size of 128 bits and a block size of 64 bits, the algorithm employs Linear Feedback Shift Registers (LFSR) for key generation. The algorithm's security is evaluated by examining its adherence to the avalanche criterion, a critical property for robust encryption. A comprehensive comparison is made between the proposed algorithm and a previously developed lightweight encryption algorithm, which also utilized LFSR for key generation in terms of avalanche effect, execution time and energy consumption. The proposed cipher is found to be significantly faster with an encryption time of 88 microseconds on Android Uno. The proposed cipher is also faster and consumed less energy in ESP 32, STM Nucleo, and Android Nano.
- 28 November 2024
- Published by Wiley in Internet Technology Letters
- Vol. 8 (5)
LiteEncrypt : A Lightweight Block Cipher for Secure Communication in IoT Enabled Sensor
The resource‐constrained characteristics of the Internet of Things (IoT) sensors lead to a pressing security requirement for developing lightweight cryptographic algorithms. This work introduces and analyzes a novel lightweight cryptographic algorithm for resource‐constrained devices. The algorithm is based on the Feistel structure in which the round function combines circular shifts, XOR, and key addition operation for data encryption and decryption. With a key size of 128 bits and a block size of 64 bits, the algorithm employs Linear Feedback Shift Registers (LFSR) for key generation. The algorithm's security is evaluated by examining its adherence to the avalanche criterion, a critical property for robust encryption. A comprehensive comparison is made between the proposed algorithm and a previously developed lightweight encryption algorithm, which also utilized LFSR for key generation in terms of avalanche effect, execution time and energy consumption. The proposed cipher is found to be significantly faster with an encryption time of 88 microseconds on Android Uno. The proposed cipher is also faster and consumed less energy in ESP 32, STM Nucleo, and Android Nano.
- 30 July 2024
- Published by Springer Nature
- p. 306-314
Intelligent IoT Based Smart Farming Framework for Soil Moisture Analysis
- 30 July 2024
- Published by Springer Nature
- p. 306-314
Intelligent IoT Based Smart Farming Framework for Soil Moisture Analysis
- Open Access
- 11 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 83140-83153
A Self-Operational Convolutional Neural Networks With Convergent Cross-Mapping and Its Application in Parkinson’s Disease Classification
Parkinson’s disease (PD) is a progressive neurodegenerative disease with multiple motor and non-motor characteristics. PD patients commonly face vocal impairments during the early stages of the disease. Therefore, diagnosis systems based on vocal disorders are at the forefront of recent PD detection studies. Our study proposes two frameworks based on Convolutional Neural Networks to classify Parkinson’s disease (PD). In recent years, Convolutional Neural Networks (CNNs) have proven highly effective in various medical applications, particularly disease classification. However, standard CNN designs have significant limitations because they require extensive manual calibration and supervision, which can result in biases and poor performance in practical applications. This paper proposes the Self-Operating Convolutional Neural Network (SOCNN) in conjunction with Convergent Cross-Mapping (CCM) to address these issues. The SOCNN architecture is intended to modify its internal parameters automatically, eliminating the need for manual intervention during training and increasing the model’s adaptability to unknown data. Adopting CCM principles, we construct a seamless connection between the input and output domains, allowing for rapid information transfer and preservation, which are crucial for accurate disease classification. To this end, we construct causal networks, extract network features, and perform deep learning analysis to distinguish Parkinson’s disease patients (PD) from age and gender-matched healthy controls (HC). Using a large dataset of Parkinson’s Disease (PD) patients and healthy controls, the effectiveness of the proposed SOCNN with CCM is evaluated. Specifically, we use the SOCNN-CCM to compute the centrality of the network nodes, which act as features for the classification models. Extensive experiments are conducted to compare the SOCNN to conventional CNN models and innovative techniques. The results demonstrate that the SOCNN-CMM outperforms state-of-the-art in terms of accuracy, sensitivity, and specificity when classifying Parkinson’s patients, confirming its diagnostic potential.
- 11 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 83140-83153
Open AccessA Self-Operational Convolutional Neural Networks With Convergent Cross-Mapping and Its Application in Parkinson’s Disease Classification
Parkinson’s disease (PD) is a progressive neurodegenerative disease with multiple motor and non-motor characteristics. PD patients commonly face vocal impairments during the early stages of the disease. Therefore, diagnosis systems based on vocal disorders are at the forefront of recent PD detection studies. Our study proposes two frameworks based on Convolutional Neural Networks to classify Parkinson’s disease (PD). In recent years, Convolutional Neural Networks (CNNs) have proven highly effective in various medical applications, particularly disease classification. However, standard CNN designs have significant limitations because they require extensive manual calibration and supervision, which can result in biases and poor performance in practical applications. This paper proposes the Self-Operating Convolutional Neural Network (SOCNN) in conjunction with Convergent Cross-Mapping (CCM) to address these issues. The SOCNN architecture is intended to modify its internal parameters automatically, eliminating the need for manual intervention during training and increasing the model’s adaptability to unknown data. Adopting CCM principles, we construct a seamless connection between the input and output domains, allowing for rapid information transfer and preservation, which are crucial for accurate disease classification. To this end, we construct causal networks, extract network features, and perform deep learning analysis to distinguish Parkinson’s disease patients (PD) from age and gender-matched healthy controls (HC). Using a large dataset of Parkinson’s Disease (PD) patients and healthy controls, the effectiveness of the proposed SOCNN with CCM is evaluated. Specifically, we use the SOCNN-CCM to compute the centrality of the network nodes, which act as features for the classification models. Extensive experiments are conducted to compare the SOCNN to conventional CNN models and innovative techniques. The results demonstrate that the SOCNN-CMM outperforms state-of-the-art in terms of accuracy, sensitivity, and specificity when classifying Parkinson’s patients, confirming its diagnostic potential.
- Open Access
- 3 May 2024
- Published by Elsevier in ICT Express
- Vol. 10 (3) , 459-464
Routing attack induced anomaly detection in IoT network using RBM-LSTM
- 3 May 2024
- Published by Elsevier in ICT Express
- Vol. 10 (3) , 459-464
Open AccessRouting attack induced anomaly detection in IoT network using RBM-LSTM
- 21 November 2023
- Published by Springer Nature
- p. 665-675
Investigating Vulnerabilities of Information Solicitation Process in RPL-Based IoT Networks
- 21 November 2023
- Published by Springer Nature
- p. 665-675
Investigating Vulnerabilities of Information Solicitation Process in RPL-Based IoT Networks
- 1 June 2023
- Published by Springer Nature
- p. 49-61
Comparative Analysis of Machine Learning Models for Customer Segmentation
- 1 June 2023
- Published by Springer Nature
- p. 49-61
Comparative Analysis of Machine Learning Models for Customer Segmentation
- 23 August 2022
- Published by Springer Nature in Cluster Computing
- Vol. 25 (5) , 3765
Correction: Mitigating the worst parent attack in RPL based internet of things
- 23 August 2022
- Published by Springer Nature in Cluster Computing
- Vol. 25 (5) , 3765
Correction: Mitigating the worst parent attack in RPL based internet of things
- 4 April 2022
- Published by Springer Nature
- p. 115-127
Partitioning Attacks Against RPL in the Internet of Things Environment
- 4 April 2022
- Published by Springer Nature
- p. 115-127
Partitioning Attacks Against RPL in the Internet of Things Environment
- 15 January 2022
- Published by Springer Nature in Cluster Computing
- Vol. 25 (2) , 1303-1320
Mitigating the worst parent attack in RPL based internet of things
- 15 January 2022
- Published by Springer Nature in Cluster Computing
- Vol. 25 (2) , 1303-1320
Mitigating the worst parent attack in RPL based internet of things