Federated Learning (FL) is a method to train machine learning (ML) models in a decentralised manner, while preserving the privacy of data from multiple clients. However, FL is vulnerable to malicious attacks, such as poisoning attacks, and is challenged by the GDPR’s “right to be forgotten”. This paper introduces a negative gradient-based machine learning technique to address these issues. Experiments on the MNIST dataset show that subtracting local model parameters can remove the influence of the respective training data on the global model and consequently “unlearn” the model in the FL paradigm. Although the performance of the resulting global model decreases, the proposed technique maintains the validation accuracy of the model above 90%. This impact on performance is acceptable for an FL model. It is important to note that the experimental work carried out demonstrates that in application areas where data deletion in ML is a necessity, this approach represents a significant advancement in the development of secure and robust FL systems.
Lighting conditions in daytime environments can reduce the object recognition rate by causing blurring, over-exposure, and shadows that mask important information about the object's shape and size. These phenomena also decrease the quality of image data, with outdoor quality being significantly lower than indoor quality. As deep learning-based object recognition algorithms heavily rely on image quality, a preprocessing process is required to improve the quality of learning image data and achieve high performance. To address this, the paper proposes a contrast-enhanced image generation preprocessing process that can improve image quality and mitigate the effects of poor lighting conditions.
The effective management of digital evidence is critical to modern forensic investigations. However, traditional evidence management approaches are often prone to security and integrity issues. In recent years, the use of blockchain technology has emerged as a promising solution to enhance the security, transparency, and integrity of digital evidence. This systematic review critically evaluates the current state of research on blockchain-based chain of custody for digital evidence and its potential to transform the digital forensic community. By analysing papers from major databases, this study provides a bibliometric analysis of the research trends and opportunities for blockchain-based evidence management since 2015. The review highlights the benefits of blockchain technology in providing an immutable and decentralised structure for documenting and auditing evidence trails. Additionally, this research identifies the challenges and limitations of implementing a blockchain-based chain of custody and presents practical and scalable solutions for overcoming these challenges at Big Data scale.
In order to connect heterogeneous nodes, objects, and smart devices of a network, such as e-transportation, e-health, e-education, e-home, and e-grip, the Internet of Things (IoT) has emerged as an efficient technology. This technology makes things easier, safer, and more productive for us all. These nodes are often resource-constrained because of their involvement in a huge network of heterogeneous devices, making them the weakest link in the chain for a cyber attacker because they generate enormous amounts of data despite a number of limitations, including memory, power, and low processor of the device. So these limitations make IoT devices vulnerable to a variety of security attacks. In this paper, we presented a survey on attacks IoT countermeasures, systematic reviews, and analyses of various IoT attacks that are occurring, classified them, discussed their defences, and identified the most significant IoT attacks countermeasures. A state-of-the-art analysis of the different attacks, including their effectiveness and degree of damage in IoT devices, has been given and contrasted. We identify the advantages and disadvantages of IoT Attack Countermeasures and proposed a Novel IoT Attack Countermeasures. Finally, we identify the open-research issue in the domain and provide directions for future research.
Forest fire prediction is an important aspect of combating forest fires. This research focuses on the effectiveness of multi-source data (lightning, hydrometric and weather) in the probability prediction of forest fires using deep learning. The results showed that the weather model had the best predictive power (average $$F1 Score = 0.955$$ ). The lightning model had an average $$F1 Score = 0.924$$ , while the hydrometric model had an average $$F1 Score = 0.690$$ . The single-source models were then merged to see the impact of the multi-source data. The multi-source model had an average $$F1 Score = 0.929$$ , whereas the average F1Score for the previous three single-source model was 0.856. The results showed that the multi-source model performed similarly to the best-performing single-source model (weather) with a 60% reduction in training data. The multi-source model had a negligible impact from the poor-performing single-source model (hydrometric).
