Android is one of the most popular mobile operating system (OS). Android is currently developed by the Google and other developers. It is an open source OS, based on the Linux kernel. Android can be used on smartphones, tablets, smartwatches and other devices. Android supports different types of applications, hardwares and firmwares. The Android operating system has different flavors and versions due to its open nature. This can lead to security and privacy concerns. Google Play Store is the most popular online app store for Android. As of 2015, more than 1.6 million applications have been listed on the Google Play Store [1]. In this project, I develop 'SecuFone: Android Security Advisor' an Android application. SecuFone evaluates the security of the overall system and provides advisory (recommendation) to improve their security. It evaluates security by using a hierarchical system. Hierarchical system uses fuzzy logic to provide users with the generalized score based on the metrics. The recommendations and score generated by the application helps users to enhance their privacy and security of their devices.

A document by Claude Carlet . Click on the document to view its contents.

Identifying cells in microscopic images is a crucial step toward studying image-based cell biology research. Cell instance segmentation provides an opportunity to study the shape, structure, form, and size of cells. Deep learning approaches for cell instance segmentation rely on the instance segmentation mask for each cell, which is a labor-intensive and expensive task. An ample amount of unlabeled microscopic data is available in the cell biology domain, but due to the tedious and exorbitant nature of the annotations needed for the cell instance segmentation approaches, the full potential of the data is not explored. This paper presents a weakly supervised approach, which can perform cell instance segmentation by using only point and bounding box-based annotation. This enormously reduces the annotation efforts. The proposed approach is evaluated on a benchmark dataset i.e., LIVECell, whereby only using a bounding box and randomly generated points on each cell, it achieved the mean average precision score of 43.53% which is as good as the full supervised segmentation method trained with complete segmentation mask. In addition, it is 3.71 times faster to annotate with a bounding box and point in comparison to full mask annotation.

This paper explores the significance of Quality of Transmission (QoT) estimation in optical networks and highlights the increasing use of machine learning (ML) techniques to enhance QoT estimation accuracy. It presents a survey of literature in this area, categorizing studies into classification and regression algorithms. ML methods are shown to improve QoT estimation, mitigate nonlinearities, and optimize decisionmaking processes. Ultimately, these advancements reduce the reliance on conservative margins, maximize network capacity, and decrease infrastructure investment. Accurate and real-time QoT information is the foundation for efficient routing and spectral allocation (RSA) systems, it enables proactive failure management, facilitates network reconfiguration, provides inputs for optical line optimization and drives optical network automation.  This paper explores and provides insights into both traditional ML models techniques and large language models (LLMs) as applied to optical communications.

This paper introduces Continual Learning for Multilingual ASR (CL-MASR), a benchmark for continual learning applied to multilingual ASR. CL-MASR offers a curated selection of medium/low-resource languages, a modular and flexible platform for executing and evaluating various CL methods on top of existing large-scale pretrained multilingual ASR models such as Whisper and AWavLM, and a standardized set of evaluation metrics.

The upcoming sixth generation (6G) is expected to support a wide range of applications that require efficient sensing, accurate localization, and reliable communication capabilities. Furthermore, 6G is expected to catalyze the development of new use cases that will require working in extreme environmental and hazardous conditions and have ultra-small size and low-cost wireless devices. Thus, developing sustainable multi-functional wireless networks that are capable of incorporating billions of low-power devices and supporting their sensing and communication requirements on top of energy harvesting capability is of paramount importance. Motivated by this, we consider in this work a rate-splitting multiple access (RSMA)-based multifunctional wireless network with sensing, energy harvesting, and communication capabilities. We employ trust region policy optimization (TRPO), a deep  reinforcement learning (DRL) algorithm, to efficiently allocate the available resources and manage the interference between the three functionalities. TRPO/DRL is capable to learn a near-optimal policy for the resource allocation problem in a complex and dynamic environment. This enables us to obtain near-optimal transmit precoders, power splitting ratios, and ratesplitting among the common and private rates in a multiple access setting. Simulation results demonstrate the effectiveness of RSMA in mitigating the interference in such multi-functional networks and its capability to accommodate the rate and energy harvesting requirements of the devices while still capable of sensing multiple targets.

A document by Van-Vi Vo . Click on the document to view its contents.

In four participants with upper limb loss, we characterized the quality and location of sensation elicited through electrical stimulation of regenerative peripheral nerve interfaces (RPNIs) over time (up to 54 months post implantation of wires). We also measured sensory perception and discomfort thresholds, sensitivity to changes in stimulation amplitude, and ability to differentiate objects of different stiffness and sizes.

In contemporary heavy-load distribution networks, preceding feasibility assessment is imperative before incorporating additional energy units. However, the feasibility examination for massive combined operational scenarios of relevant units is computationally intensive with repetitive power flow calculations. To this end, this paper proposes a rapid assessment framework, the kernel of which is to learn from formerly examined scenarios, thus forming expansive feasible/infeasible regions to geometrically rule in/out subsequent scenarios. By sidestepping the power flow computation in most scenarios, we accelerate the assessment process. Furthermore, enlightened by heuristic hypersurface search, such prechecking efficiency can be boosted by resorting the scenario sequence. In a risk-averse manner, this framework can be conceptualized using the exact grid model, though it is initially designed for the convex grid model. Evidenced by testing on two different-scale distribution grids, the framework shows a significant assessment efficiency improvement and strict accuracy guarantee, where we observe at least 76.13% assessment time reduction and zero accuracy loss. We anticipate this work to be a starting point for more sophisticated geometry-accelerating feasibility assessment methods.

This research paper considers a dataset of students of a particular class and uses different regression algorithms to predict an overall rating for the students. These regression algorithms are then compared with each other and further analyzed according to their performances.

We propose a novel paradigm for cone-beam computed tomography (CBCT) reconstruction from ultra-sparse X-ray projections, by introducing a framework that generates auxiliary X-ray projections under controlled geometric parameters. This innovation overcomes the limitations of conventional methods that are constrained to producing fixed-angle projections. Our approach is organized into three key modules: the XGen, X-Correction, and CT-Correction module. Through the XGen module, we generate projections based on any given geometric parameters to supplement the geometric information in the projection domain. The X-Correction module then introduces geometric corrections to harmonize the generated projections. Finally, through the CT-Correction module, the reconstructed image undergoes refining, thereby enhancing the image quality within the image domain. We have validated our model on several datasets including a large-scale publicly available lung CT dataset (LIDC-IDRI with 1018 patients); an extensive abdominal CT dataset (AbdomenCT-1K, with a selected 1k patients); and our proprietary pelvic CT dataset, collated from a hospital (445 patients). Real walnut projection data were also incorporated for genuine projection validation. Compared to the traditional projection generation methods and the state-of-the-art ultra-sparse reconstruction techniques on 2-view and 10-view tasks, our method has demonstrated consistently superior performance across various tasks.

Visual grounding of remote sensing (RSVG) aims to detect the specific objects associated with query expressions in remote sensing (RS) images.

This study presents the first automated robotic system for PSP without the need for an external tracking system. The proposed system utilizes a U-Net-based framework to segment 2d US images and reconstruct 3D anatomy features, enabling visualization of the spinal structure. Then, the desired screw trajectories are automatically registered to the 3D reconstruction. By employing this robot-assisted system, the reliance on guide-tube-based approaches is eliminated, reducing the physical demands of the procedure and allowing the surgeon to focus more on planning and evaluation rather than execution. Experimental evaluation on two lamb vertebrae yielded four screw trajectories. The mean 3D error is 2.78 mm at the entry points and 4.82 mm at the stop points. Compared to commercial robotic systems, the use of non-radiation imaging decreases the second harm to the patient and surgeons. Meanwhile, it gets rid of the bone pin, which can lead to an incision in the patient’s body.

A document by Andrea Detti. Click on the document to view its contents.

In this article, a compact 3D beam-steering MIMO antenna operating at 5.75 GHz is proposed for on-body Internet of Things (IoT) applications. The planar antenna comprises a central patch and two concentric annular rings. The theory of spherical modes is used to discuss the beam-steering concept, where the free space measured beam-scanning covers the entire xy-plane, while a 44° range is realized in xz and yz-planes. The on-body performance is tested using a multi-layer phantom, and the measured peak realized gain is 6.41 dBi. The antenna is characterized by compact size (0.77λ), low profile (0.03λ), specific absorption rate below the established limits, good efficiency and 3D beam-steering characteristics while operating in an on-body setup. Therefore, the proposed solution can enable advanced wireless applications like localization and physical layer security in emerging size-constrained on-body IoT devices.

Assessment of total transfer capability (TTC) is vital for determining the permissible power transfer between two areas of an interconnected power system. In the context of heightened volatility and time-variability in power system operating states after integrating high proportions of renewable energy, data-driven inferential assessment methods emerge as promising alternatives, offering faster assessment capabilities compared to knowledge-based iterative methods. However, data-driven methods typically struggle to establish reliable connections between assessment outcomes and security standards, hindering the guarantee of conservatism. A hybrid algorithm, combining knowledge-based and data-driven techniques, is proposed to accurately and efficiently assess TTC while strictly complying with pre-established security and stability constraints. Data-driven inference accelerates knowledge-based iterative processes by rapidly identifying reasonable initial values and providing adaptive step sizes, while knowledge-based analysis guides data-driven methods through offering stability margin information. This mechanism leverages the speed of data-driven methods while maintaining conservatism through knowledge-based approaches. The effectiveness of the proposed method is verified on benchmarks, including the IEEE 30-bus system and a real-world power system, which also exhibits conservatism and robustness in the face of increasing renewable energy penetration.

Demand for autonomous protection in computing devices can not go unnoticed with a cataclysmic rise in cyber-attacks. Consequently, cybersecurity measures with an improved generalization that can proactively determine the indicators of compromises to predict zero-day threats or previously unseen malware together with known malware are highly desirable. In this article, we present a novel concept of autonomous device protection based on behavioral profiling by continuously monitoring internal resource usage and exploiting a large language model to distinguish between benign and malicious behavior. We design and develop a proof-of-concept for Windows-based computing devices relying on a built-in event tracing mechanism for log collection that is converted into structured data using a graph data structure. We extract graph-level features, \textit{i.e., graph depth, nodes count, number of leaf nodes, node degree statistics, and events count}, and node-level features, i.e., process start, file create and registry events details for each graph. Further, we exploit a pre-trained large language network - a simple contrastive sentence embedding framework to extract strong features, i.e., dense vectors, from event graphs. Finally, we train a random forest classifier using both the graph- and node-level features to obtain classification models that are evaluated on a collected dataset containing one thousand benign and malicious samples achieving accuracy up to 99.25%.

Ransomware attacks are becoming one of the most disruptive threats in modern cybersecurity, causing significant financial and operational damage to organizations worldwide. Addressing the limitations of traditional detection techniques, a novel cooperative clustering framework has been introduced, leveraging machine learning to enable real-time collaboration between multiple detection agents across distributed network environments. The system excels in identifying both known and emerging ransomware variants through unsupervised learning algorithms that autonomously group anomalous behaviors while maintaining low false positive rates. Extensive experiments demonstrated that the proposed framework outperforms existing methods in terms of detection accuracy, processing time, and scalability, even under varying network loads and attack complexities. The integration of real-time collaboration between detection agents not only increases resilience against sophisticated ransomware attacks but also ensures a rapid, autonomous response to threats, minimizing the overall impact on the targeted infrastructure. These findings highlight the potential for cooperative clustering to significantly enhance the effectiveness of ransomware detection and mitigation within broader cybersecurity frameworks, offering a scalable and adaptive defense solution.