Negalign Wake Hundera

Large Language Models (LLMs) have revolutionized natural language processing, yet their deployment is hindered by data, computation, and privacy constraints. Federated Learning (FL) offers a promising solution by enabling collaborative, privacy-preserving training across distributed devices, while the push for low-latency on-device intelligence further drives LLM integration into FL and edge settings—posing new challenges in heterogeneity and resource limits. This survey comprehensively reviews the integration of LLMs with federated learning, termed FLM, and its deployment at the edge, with particular emphasis on the robustness, privacy, and trustworthiness challenges that emerge across the LLM lifecycle from pre-training to deployment. We analyze core challenges including communication cost, system heterogeneity, privacy risks, and scalability, with a focus on edge-oriented efficiency techniques such as pruning and quantization. Security vulnerabilities and defenses are also discussed, alongside trade-offs among privacy, robustness, and performance. We further examine demographic, contribution-related, and performance-related biases that can emerge in FLM systems. Finally, we outline open research directions, underscoring the potential of federated and edge intelligence to enable scalable, privacy-aware LLM ecosystems, and aim to offer a unified perspective to guide future research in this fast-moving field.

Wireless body area networks (WBANs) enable continuous and remote monitoring of medical information in modern healthcare systems. However, ensuring the security and privacy of WBAN data remains a significant challenge, especially with the rise of algorithm substitution attacks (ASAs) that enable adversaries to perform filtering attacks during data sharing. To address these challenges, we introduce a novel blockchain-based heterogeneous certificateless online/offline signcryption scheme with proxy re-encryption and a cryptographic reverse firewall (BHCOSP-CRF). The scheme establishes secure data exchange between a sender using a certificateless cryptosystem and a receiver operating in an identity-based cryptosystem, with a CRF integrated to protect against ASAs. To reduce the computational overload on WBAN nodes, BHCOSP-CRF divides signcryption into offline and online phases. Most computationally intensive operations are performed in the offline phase without the message, whereas only lightweight computations are executed in the online phase once the message becomes available. Furthermore, a consortium blockchain service center (CBSC) stores the initial ciphertext and re-encrypts it via smart contracts, which avoids single points of failure and improves efficiency compared to traditional proxy servers. Moreover, we prove the security of BHCOSP-CRF in the random oracle model, and performance evaluation shows that BHCOSP-CRF achieves better performance than existing schemes in terms of communication cost, computational efficiency, energy consumption, and latency. Therefore, BHCOSP-CRF is well-suited for WBAN environments.

In the context of smart cities, sensors are pivotal for the collection and analysis of health data within smart healthcare systems. Despite their importance, securing this data for cloud storage poses significant challenges, particularly in terms of data searchability and encryption. Our paper introduces a novel solution, namely certificateless encryption supporting equality test with cryptographic reverse firewalls (CLE-ET-CRF). This protocol allows cloud servers to perform equality tests on encrypted data without compromising its confidentiality, effectively mitigating risks like offline message recovery attacks (OMRA) and algorithm substitution attacks (ASAs). By eliminating the need for traditional certificate management and addressing key escrow concerns, CLE-ET-CRF provides a secure, efficient, and privacy-preserving mechanism for managing healthcare data in smart city ecosystems. Our evaluation confirms that the protocol meets the high standards required for privacy and efficiency in smart healthcare applications.

With the rapid growth of Internet of Things (IoT) applications and decentralized data-driven systems, secure and flexible data sharing remains a critical challenge. Identity-Based Proxy Re-Encryption (IB-PRE) is an effective cryptographic primitive for enabling fine-grained access delegation without exposing plaintext data. However, existing IB-PRE schemes remain vulnerable to algorithm substitution attacks (ASA), malicious key generation, inefficient ciphertext management, and the lack of practical revocation mechanisms. To address these limitations, we propose a blockchain-based revocable identity-based proxy re-encryption scheme with cryptographic reverse firewalls (BRIBPR-CRF) for secure IoT data sharing. The proposed scheme integrates CRFs into the key generation, encryption, and proxy re-encryption key generation processes to mitigate ASA and limit the impact of potentially compromised cryptographic components. An efficient identity revocation mechanism enables the removal of compromised or expired users without requiring global key updates. In addition, a consortium blockchain is employed to record initial ciphertexts and execute proxy re-encryption via smart contracts, thereby eliminating single points of failure and reducing trust assumptions compared with traditional proxy-based systems. We formally prove the security of BRIBPR-CRF in the random-oracle model. Finally, extensive performance evaluations demonstrate that the proposed scheme achieves lower communication, computation, and energy overhead compared with existing schemes, making it suitable for secure and scalable data sharing in decentralized and semi-trusted IoT environments.

With the development of the Industrial Internet of Things (IIoT), the amount of data generated by industrial manufacturing equipment is growing rapidly, creating a critical need for secure and efficient cloud-based data sharing. While bilateral access control enables both data senders and receivers to define their own fine-grained access policies, existing schemes transmit these policies in plaintext, exposing sensitive operational metadata and creating security vulnerabilities. Furthermore, they lack essential functionalities for practical IIoT deployment, including efficient duplicate data detection, a robust user revocation mechanism, and computationally lightweight operations suitable for resource-constrained devices. To address these challenges, this article proposes a novel policy-hiding, lightweight, and revocable bilateral access control scheme with equality test (RBAC-ET) for cloud-enabled IIoT systems. RBAC-ET is the first scheme to integrally combine five critical capabilities: first, fine-grained bilateral access control using linear secret sharing schemes (LSSSs) supporting and/or logical operations. Second, policy confidentiality is achieved by encrypting access policies within the ciphertext to prevent the leakage of sensitive relationships. Third, the equality test functionality enables cloud servers to identify duplicate ciphertexts for efficient storage and processing without requiring decryption. Fourth, a proposed lightweight design is based on elliptic curve scalar multiplication operations instead of the more resource-intensive bilinear pairing operations. Fifth, a robust user revocation mechanism that ensures both forward and backward secrecy in dynamic industrial environments. Theoretical analysis and experimental results demonstrate that RBAC-ET offers superior security and functionality while maintaining computational efficiency comparable to that of state-of-the-art schemes, making it a scalable and practical solution for modern IIoT data-sharing applications.

Today, glaucoma detection is a severe problem. Because it was predicted that there would be 111.8 million cases of glaucoma worldwide by 2040, up from the expected 76.0 million cases in 2020. Blindness results if early treatment is not given. The CNN model has been used in a variety of studies to identify glaucoma. The results obtained thus far are insufficient to reliably identify glaucoma. In order to detect glaucoma, which is a binary classification, we are utilizing the CNN model. We divided the REFUGUE data set into tanning and testing, with 80% of the photos being utilized for tanning and 20% for testing. Our model's accuracy score was 99.8% for the training set and 95.4% for the testing set. We get result of precision=98% and recall= 97%. We compare the model with other state of arts (SOTA) our experimental results reveal that, the proposed model for glaucoma detection using convolutional neural network is more proficient in glaucoma classification.

To meet the increasing demand for its services, a cloud system should make optimum use of its available resources. Additionally, the high and low oscillations in cloud workload are another significant symmetrical issue that necessitates consideration. A suggested particle swarm optimization (PSO)-based ensemble meta-learning workload forecasting approach uses base models and the PSO-optimized weights of their network inputs. The proposed model employs a blended ensemble learning strategy to merge three recurrent neural networks (RNNs), followed by a dense neural network layer. The CPU utilization of GWA-T-12 and PlanetLab traces is used to assess the method’s efficacy. In terms of RMSE, the approach is compared to the LSTM, GRU, and BiLSTM sub-models.