Jishu Kinkar Medhi

Mr. Jishu Kinkar Medhi

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
31 May 2025 in Neural Computing and Applications
Verified
Journal Article
Open access
PDF file
JMJishu Kinkar Medhi
Jishu Kinkar Medhi
RLRui Liu
Rui Liu
QWQun Wang
Qun Wang
XCXuhui Chen
Xuhui Chen
The growing adoption of unmanned aerial vehicles (UAV) in civilian applications has highlighted the necessity of securing these systems against potential cyber threats. This paper presents a distilled pruned network for unmanned aerial vehicles (UAV-DiPNID), a specialized intrusion detection system designed to monitor and detect unauthorized or malicious activities involving UAVs. The proposed method leverages a combination of knowledge distillation and model pruning to develop an effective and efficient intrusion detection system based on a deep neural network. Knowledge distillation transfers knowledge from a larger, more complex model (teacher) to a smaller, simpler model (student) to enhance inference time and model compactness. Additionally, model pruning is employed to reduce the model’s size by removing unnecessary connections and weights, further optimizing resource usage. The experimentation and evaluation of the proposed method demonstrate its effectiveness in terms of intrusion detection accuracy, inference time, and model size. Extensive experiments on the UAV-IDS 2020 dataset demonstrate that UAV-DiPNID achieves an average intrusion detection accuracy of 99.61%, outperforming the state-of-the-art benchmark of 99.37%. Furthermore, our method reduces the inference time by up to 80.70% and the model size by up to 90% compared to conventional convolutional neural network-based approaches. These results highlight the superiority of UAV-DiPNID compared to baseline intrusion detection systems for UAVs, making it well-suited for real-time applications in resource-constrained environments. Furthermore, the accuracy of intrusion detection remains high, showcasing the model’s robustness in capturing complex patterns and features.

19 November 2023 in Information
Verified
Journal Article
Open access
PDF file
JMJishu Kinkar Medhi
Jishu Kinkar Medhi
RLRui Liu
Rui Liu
Qianlong WangQianlong Wang
Qianlong Wang
XCXuhui Chen
Xuhui Chen
Multiple unmanned aerial vehicle (multi-UAV) systems have gained significant attention in applications, such as aerial surveillance and search and rescue missions. With the recent development of state-of-the-art multiagent reinforcement learning (MARL) algorithms, it is possible to train multi-UAV systems in collaborative and competitive environments. However, the inherent vulnerabilities of multiagent systems pose significant privacy and security risks when deploying general and conventional MARL algorithms. The presence of even a single Byzantine adversary within the system can severely degrade the learning performance of UAV agents. This work proposes a Byzantine-resilient MARL algorithm that leverages a combination of geometric median consensus and a robust state update model to mitigate, or even eliminate, the influence of Byzantine attacks. To validate its effectiveness and feasibility, the authors include a multi-UAV threat model, provide a guarantee of robustness, and investigate key attack parameters for multiple UAV navigation scenarios. Results from the experiments show that the average rewards during a Byzantine attack increased by up to 60% for the cooperative navigation scenario compared with conventional MARL techniques. The learning rewards generated by the baseline algorithms could not converge during training under these attacks, while the proposed method effectively converged to an optimal solution, proving its viability and correctness.

7 November 2023 in Mathematics
Verified
Journal Article
Open access
PDF file
JMJishu Kinkar Medhi
Jishu Kinkar Medhi
PRPusheng Ren
Pusheng Ren
MHMengsha Hu
Mengsha Hu
XCXuhui Chen
Xuhui Chen
Deep learning is a promising technique for bioelectrical signal analysis, as it can automatically discover hidden features from raw data without substantial domain knowledge. However, training a deep neural network requires a vast amount of labeled samples. Additionally, a well-trained model may be sensitive to the study object, and its performance may deteriorate sharply when transferred to other study objects. We propose a deep multi-task learning approach for bioelectrical signal analysis to address these issues. Explicitly, we define two distinct scenarios, the consistent source-target scenario and the inconsistent source-target scenario based on the motivation and purpose of the tasks. For each scenario, we present methods to decompose the original task and dataset into multiple subtasks and sub-datasets. Correspondingly, we design the generic deep parameter-sharing neural networks to solve the multi-task learning problem and illustrate the details of implementation with one-dimension convolutional neural networks (1D CNN), vanilla recurrent neural networks (RNN), recurrent neural networks with long short-term memory units (LSTM), and recurrent neural networks with gated recurrent units (GRU). In these two scenarios, we conducted extensive experiments on four electrocardiogram (ECG) databases. The results demonstrate the benefits of our approach, showing that our proposed method can improve the accuracy of ECG data analysis (up to 5.2%) in the MIT-BIH arrhythmia database.

6 October 2025 in IEEE Transactions on Network Science and Engineering
Verified
Journal Article
Closed access
JMJishu Kinkar Medhi
Jishu Kinkar Medhi
XCXuhui Chen
Xuhui Chen
Miao PanMiao Pan
Miao Pan
PLPan Li
Pan Li
The Internet of Things (IoT) emerges as a ubiquitous information collection and processing paradigm that can potentially exploit massive data for various applications using machine learning technologies. However, these data are usually unlabeled, and the labeling process is usually both time and effort consuming. While active learning can reduce data labeling costs, the two computation-intensive operations, i.e., sample selection and learning model training, hinder the use of active learning on resource-limited IoT devices. Although cloud computing may relieve users from expensive computations, directly outsourcing data and computations to the cloud raises critical security and privacy concerns. In this paper, we develop a secure outsourcing framework for deep active learning (SEDAL) by considering a general active learning framework with a deep neural network (DNN) learning model. Instead of using computationally expensive encryption based methods, we utilize a lightweight linear algebra based secure matrix transformation scheme to protect the user's input data and the DNN model. Compared to traditional homomorphic encryption based secure outsourcing schemes, our scheme reduces the computational complexity at the user from $O(n^{3})$ to $O(n^{2})$. We evaluate the performance by implementing it on an arrhythmia diagnosis application. Experimental results show that the proposed scheme can obtain a well-trained classifier using fewer queried samples, the computation time outperforms the traditional homomorphic encryption and the state-of-the-art privacy-preserving outsourcing scheme (EVPP), and the computation time and communication overheads are acceptable and practical.