External Server Research Articles

The Devil is in the Details: Detection, Measurement and Lawfulness of Server-Side Tracking on the Web

As online privacy is cementing itself as one of the core pillars of the Internet, major changes are happening across many industries. On the technological side, users are pushing for more privacy-preserving technologies and rely on browsers and extensions that limit online tracking as much as possible. On the legal front, regulations like GDPR and the ePrivacy Directive in Europe have forced companies to change their practices and be more transparent about how they handle user data. For the ad industry, the end of third-party cookies planned for 2025 is having severe ramifications as the main source of data on which this industry is built on will be gone. In this tumultuous context, companies have come up with innovative ways to overcome current and future restrictions. A novel technique which has not received much attention called Server-side tracking (SST) moves its tracking logic away from the user's device onto an external server. In this work, our aim is to detect SST on the web and understand its lawfulness with respect to current legislation. We developed a methodology that relies on crawls spaced 2 years apart performed before and after the introduction of SST to identify trackers that moved behind SST domains and that are now hidden from view. Our results show that 389, out of 7,367 visited websites, track users behind a cloaked domain and that 28 websites perform Server-side tracking in a first-party capacity. We demonstrate that such a tracking technique can overcome the Same-Origin Policy and introduce security vulnerabilities. Together with a legal scholar, we also show that SST entails non-compliant practices and infringes the GDPR and the ePrivacy Directive.

Accelerating QKD post-processing by secure offloading of information reconciliation

While quantum key distribution (QKD) offers unparalleled security in communication, its real-world application is hindered by inherent physical constraints. The challenge lies predominantly in the cumbersome, energy-intensive nature of current QKD systems, which stems largely from the time-intensive post-processing stage. This paper investigates the feasibility of offloading the computationally intensive post-processing tasks, specifically focusing on information reconciliation (IR), to potentially untrusted servers.We present a novel scheme that leverages syndrome decoding techniques to efficiently transfer the IR step of QKD protocols to a single external server. Notably, this offloading is accomplished while maintaining the highest level of security, known as unconditional security. The proposed technique is bolstered by a comprehensive theoretical analysis and validated through experimental trials. These findings demonstrate the effectiveness of our approach in bridging the gap between the theoretical promise of QKD and its real-world deployment.

LLM-AIx: An open source pipeline for Information Extraction from unstructured medical text based on privacy preserving Large Language Models.

In clinical science and practice, text data, such as clinical letters or procedure reports, is stored in an unstructured way. This type of data is not a quantifiable resource for any kind of quantitative investigations and any manual review or structured information retrieval is time-consuming and costly. The capabilities of Large Language Models (LLMs) mark a paradigm shift in natural language processing and offer new possibilities for structured Information Extraction (IE) from medical free text. This protocol describes a workflow for LLM based information extraction (LLM-AIx), enabling extraction of predefined entities from unstructured text using privacy preserving LLMs. By converting unstructured clinical text into structured data, LLM-AIx addresses a critical barrier in clinical research and practice, where the efficient extraction of information is essential for improving clinical decision-making, enhancing patient outcomes, and facilitating large-scale data analysis. The protocol consists of four main processing steps: 1) Problem definition and data preparation, 2) data preprocessing, 3) LLM-based IE and 4) output evaluation. LLM-AIx allows integration on local hospital hardware without the need of transferring any patient data to external servers. As example tasks, we applied LLM-AIx for the anonymization of fictitious clinical letters from patients with pulmonary embolism. Additionally, we extracted symptoms and laterality of the pulmonary embolism of these fictitious letters. We demonstrate troubleshooting for potential problems within the pipeline with an IE on a real-world dataset, 100 pathology reports from the Cancer Genome Atlas Program (TCGA), for TNM stage extraction. LLM-AIx can be executed without any programming knowledge via an easy-to-use interface and in no more than a few minutes or hours, depending on the LLM model selected.

Hierarchical Clustering via Single and Complete Linkage Using Fully Homomorphic Encryption.

Hierarchical clustering is a widely used data analysis technique. Typically, tools for this method operate on data in its original, readable form, raising privacy concerns when a clustering task involving sensitive data that must remain confidential is outsourced to an external server. To address this issue, we developed a method that integrates Cheon-Kim-Kim-Song homomorphic encryption (HE), allowing the clustering process to be performed without revealing the raw data. In hierarchical clustering, the two nearest clusters are repeatedly merged until the desired number of clusters is reached. The proximity of clusters is evaluated using various metrics. In this study, we considered two well-known metrics: single linkage and complete linkage. Applying HE to these methods involves sorting encrypted distances, which is a resource-intensive operation. Therefore, we propose a cooperative approach in which the data owner aids the sorting process and shares a list of data positions with a computation server. Using this list, the server can determine the clustering of the data points. The proposed approach ensures secure hierarchical clustering using single and complete linkage methods without exposing the original data.

A Quasi Real-Time Evaluation of High-Resolution Mass Spectra of Complex Chlorinated Paraffin Mixtures and Their Transformation Products.

Chlorinated paraffins (CPs) are complex mixtures of polychlorinated n-alkanes with multiple carbon- (C-, nC = 9-30) and chlorine homologues (Cl-, nCl = 3-18). The mass spectrometric analysis of CPs is time-consuming and challenging, especially when interferences between CPs, their transformation products, or from the matrix are numerous. These analytical challenges and the lack of appropriate and accessible data evaluation tools are obstacles to their analysis. CP-Hunter is a web-based, open-access data processing platform for the automatic analysis of mass spectra of CPs and their transformation products. Extracts of two consumer plastic materials and sewage sludge were evaluated with CP-Hunter. C- and Cl-homologue distributions were obtained in quasi-real-time and the posterior calculated fingerprints were in agreement with the ones obtained by traditional methods. However, the data extraction and evaluation time were now reduced from several minutes to seconds. The implemented signal deconvolution method, i.e., to resolve mass spectrometric interferences, provides robust results, even when severe matrix effects are present. CP-Hunter facilitates the untargeted analysis of unknown products and the detection and elimination of false positive signals. Finally, data evaluation with CP-Hunter is performed locally without the transfer of data to external servers. The tool is safe, public, and accessible at https://cphunter.cheminfo.org/.

BSA-SGRU-A Novel Deep Learning Framework for Alleviate the Multiple Attacks in IoT-Cloud Environment

The Internet of Things (IoT) has emanated as an innovative technology that grants users to establish connections to external servers located in the cloud and the uninterrupted Internet, allowing them to live more comfortably. As the number of users increases, security and privacy breaches also increase exponentially, as it invests the greatest threats to the IoT networked devices and cloud storage. Though machine and deep learning algorithms have paved the bright light toward the design of intelligent systems to mitigate the attacks, creating effective defenses against numerous assaults remains a significant design challenge for researchers, and accurately predicting these attacks continues to pose an ongoing hurdle. To solve this problem, this paper proposes a hybrid learning model that ensembles the skip connection-based gated recurrent units (Skip-GRU) and bi-layered self-attention networks to predict and mitigate the different attacks. First, Skip-GRU is constructed using residual connections that remove redundant information and capture only global features. Second, novel bi-layered privacy-preserving networks are combined to obtain the spatial and temporal attributes. Finally, softmax is utilized to obtain the prediction results of sample labels. Performance parameters including accuracy, precision, recall, and F1-score are assessed and evaluated during the comprehensive eradication exploration that is conducted utilizing the NSL-KDD and UNSW datasets. By comparing how it performs to other cutting-edge approaches to learning, the one recommended demonstrates its superiority. The technique has proven its efficacy in providing enough security versus various threats, as seen by the results, which show that it has obtained 0.97 accuracy, 0.96 precision, 0.96 recall, and 0.96 F1-score.

Intelligent Bladder Volume Monitoring for Wearable Ultrasound Devices: Enhancing Accuracy Through Deep Learning-Based Coarse-to-Fine Shape Estimation.

Accurate and continuous bladder volume monitoring is crucial for managing urinary dysfunctions. Wearable ultrasound (US) devices offer a solution by enabling noninvasive and real-time monitoring. Previous studies have limitations in power consumption and computation cost or quantitative volume estimation capability. To alleviate this, we present a novel pipeline that effectively integrates conventional feature extraction and deep learning (DL) to achieve continuous quantitative bladder volume monitoring efficiently. Particularly, in the proposed pipeline, bladder shape is coarsely estimated by a simple bladder wall detection algorithm in wearable devices, and the bladder wall coordinates are wirelessly transferred to an external server. Subsequently, a roughly estimated bladder shape from the wall coordinates is refined in an external server with a diffusion-based model. With this approach, power consumption and computation costs on wearable devices remained low, while fully harnessing the potential of DL for accurate shape estimation. To evaluate the proposed pipeline, we collected a dataset of bladder US images and RF signals from 250 patients. By simulating data acquisition from wearable devices using the dataset, we replicated real-world scenarios and validated the proposed method within these scenarios. Experimental results exhibit superior improvements, including +9.32% of IoU value in 2-D segmentation and -22.06 of RMSE in bladder volume regression compared to state-of-the-art (SOTA) performance from alternative methods, emphasizing the potential of this approach in continuous bladder volume monitoring in clinical settings. Therefore, this study effectively bridges the gap between accurate bladder volume estimation and the practical deployment of wearable US devices, promising improved patient care and quality of life.

SoK: Wildest Dreams: Reproducible Research in Privacy-preserving Neural Network Training

Machine Learning (ML), addresses a multitude of complex issues in multiple disciplines, including social sciences, finance, and medical research. ML models require substantial computing power and are only as powerful as the data utilized. Due to the high computational cost of ML methods, data scientists frequently use Machine Learning-as-a-Service (MLaaS) to outsource computation to external servers. However, when working with private information, like financial data or health records, outsourcing the computation might result in privacy issues. Recent advances in Privacy-Preserving Techniques (PPTs) have enabled ML training and inference over protected data through the use of Privacy-Preserving Machine Learning (PPML). However, these techniques are still at a preliminary stage and their application in real-world situations is demanding. In order to comprehend the discrepancy between theoretical research suggestions and actual applications, this work examines the past and present of PPML, focusing on Homomorphic Encryption (HE) and Secure Multi-party Computation (SMPC) applied to ML. This work primarily focuses on the ML model's training phase, where maintaining user data privacy is of utmost importance. We provide a solid theoretical background that eases the understanding of current approaches and their limitations. We also provide some preliminaries of SMPC, HE, and ML. In addition, we present a systemization of knowledge of the most recent PPML frameworks for model training and provide a comprehensive comparison in terms of the unique properties and performances on standard benchmarks. Also, we reproduce the results for some of the surveyed papers and examine at what level existing works in the field provide support for open science. We believe our work serves as a valuable contribution by raising awareness about the current gap between theoretical advancements and real-world applications in PPML, specifically regarding open-source availability, reproducibility, and usability.

A short review of the ADAS progress in the last decade and the potential concept of Human-Machine Symbiosis (HMS)

Abstract This short review highlights the main sensor structures that are used in the ADAS (Advanced Driver Assisted System) field, in order to outline the progress at this moment (2023). Starting from this achieved level, we have proposed another approach to the ADAS problematics, regarding the prospects for development after the current moment, a different perspective than the one that has AV (Autonomus Vehicle) as its endpoint. Computer Assisted Driving represents a necessary and sufficient solution for increasing traffic safety in the near future. The integrated driver concept in the assisted environment implies a symbiotic human-machine collaboration. This collaboration requires a psychological acceptance of the ADAS system on behalf of humans, as an essential part of the current and future automobile. The progress highlighted by the documentation undertaken allows us to imagine a concept of a symbiotic human-machine system, based on the communication between ADAS and an external computing resource that assists the driver, based on the long analysis of recorded car events. Future automobiles must include an ADAS system that provides a subtle human-machine synapse, based on a series of computerized warnings to the driver generated in a timely manner, resulting from the high predictive capacity of AI-DLM algorithms. The computing speed of the local ADAS algorithms will be supplemented by the external server processing resource, accessed at the right time, to resolve an unexpected deadlock in the car roll.

Novel GA-Based DNN Architecture for Identifying the Failure Mode with High Accuracy and Analyzing Its Effects on the System

Symmetric data play an effective role in the risk assessment process, and, therefore, integrating symmetrical information using Failure Mode and Effects Analysis (FMEA) is essential in implementing projects with big data. This proactive approach helps to quickly identify risks and take measures to address them. However, this task is always time-consuming and costly. On the other hand, there is an essential need for an expert in this field to carry out this process manually. Therefore, in the present study, the authors propose a new methodology to automatically manage this task through a deep-learning technique. Moreover, due to the different nature of the risk data, it is not possible to consider a single neural network architecture for all of them. To overcome this problem, a Genetic Algorithm (GA) was employed to find the best architecture and hyperparameters. Finally, the risks were processed and predicted using the new proposed methodology without sending data to other servers, i.e., external servers. The results of the analysis for the first risk, i.e., latency and real-time processing, showed that using the proposed methodology can improve the detection accuracy of the failure mode by 71.52%, 54.72%, 72.47%, and 75.73% compared to the unique algorithm with the activation function of Relu and number of neurons 32, respectively, related to the one, two, three, and four hidden layers.

Take the aTrain. Introducing an interface for the Accessible Transcription of Interviews

Research in behavioral and experimental finance becomes more multifaceted and the analysis of data from speech interactions more important. This raises the need for technical support for researchers using qualitative data generated from speech interactions. aTrain serves this need and is an open-source, offline transcription tool with a graphical interface for audio data in multiple languages. It requires no programming skills, runs on most computers, operates without internet, and ensures data is not uploaded to external servers. aTrain combines OpenAI’s Whisper transcription models with speaker recognition and provides output that integrates with MAXQDA and ATLAS.ti. Available on the Microsoft Store for easy installation, its source code is also accessible on GitHub. aTrain, designed for speed on local computers, transcribes audio files at 2-3 times the audio duration on mobile CPUs using the highest-accuracy Whisper transcription models. With an entry-level graphics card, this speed improves to 30% of the audio duration.

Machine Learning for Columnar High Energy Physics Analysis

Machine learning (ML) has become an integral component of high energy physics data analyses and is likely to continue to grow in prevalence. Physicists are incorporating ML into many aspects of analysis, from using boosted decision trees to classify particle jets to using unsupervised learning to search for physics beyond the Standard Model. Since ML methods have become so widespread in analysis and these analyses need to be scaled up for HL-LHC data, neatly integrating ML training and inference into scalable analysis workflows will improve the user experience of analysis in the HL-LHC era.We present the integration of ML training and inference into the IRISHEP Analysis Grand Challenge pipeline to provide an example of how this integration can look like in a realistic analysis environment. We also utilize Open Data to ensure the project’s reach to the broader community. Different approaches for performing ML inference at analysis facilities are investigated and compared, including performing inference through external servers. Since ML techniques are applied for many different types of tasks in physics analyses, we showcase options for ML integration that can be applied to various inference needs.

FusionFlow: Accelerating Data Preprocessing for Machine Learning with CPU-GPU Cooperation

Data augmentation enhances the accuracy of DL models by diversifying training samples through a sequence of data transformations. While recent advancements in data augmentation have demonstrated remarkable efficacy, they often rely on computationally expensive and dynamic algorithms. Unfortunately, current system optimizations, primarily designed to leverage CPUs, cannot effectively support these methods due to costs and limited resource availability. To address these issues, we introduce FusionFlow, a system that cooperatively utilizes both CPUs and GPUs to accelerate the data preprocessing stage of DL training that runs the data augmentation algorithm. FusionFlow orchestrates data preprocessing tasks across CPUs and GPUs while minimizing interference with GPU-based model training. In doing so, it effectively mitigates the risk of GPU memory overflow by managing memory allocations of the tasks within the GPU-wide free space. Furthermore, FusionFlow provides a dynamic scheduling strategy for tasks with varying computational demands and reallocates compute resources on the fly to enhance training throughput for both single and multi-GPU DL jobs. Our evaluations show that FusionFlow outperforms existing CPU-based methods by 16--285% in single-machine scenarios and, to achieve similar training speeds, requires 50--60% fewer CPUs compared to utilizing scalable compute resources from external servers.

Social navigation framework for autonomous vehicle with hierarchical cyber-physical system architecture

An autonomous vehicle operating alongside humans should ideally have a high social capability, such as being able to communicate with humans, negotiate space, predict reactions, etc. This can be achieved by a prediction feature trained with diverse data on human behavior. Hierarchical cyber-physical system (CPS) architecture design, which sends the prediction feature to an external server while observing a limited operational design domain (ODD) and acquiring data continuously, has great potential to refine the training process as well as improve the performance. However, this architecture design requires the planning and prediction modules to be explicitly decoupled, which takes away from the recent success on social navigation. In this paper, we propose a novel autonomous navigation framework enabling social behavior while decoupling the planning and prediction modules to take advantage of the hierarchical CPS architecture. In the proposed framework, pedestrian trajectories are predicted as reactions to pre-generated candidates for an ego vehicle trajectory, and the ego vehicle trajectory is then selected to maximize mutual benefit to both the ego vehicle and surrounding pedestrians. We evaluated the proposed framework with simulations using the social force model and found that it was able to achieve the social behavior.

Design of exact reduct rough set hardware accelerator for early-stage diabetes risk prediction

Diabetes is a prevalent, chronic metabolic disorder worldwide. Detecting and predicting risks early are vital for timely intervention. In this paper, the design and implementation of an Exact Reduct Rough Set Hardware Accelerator is presented for early-stage diabetes risk prediction. The proposed hardware accelerator leverages the principles of Rough Set theory to identify essential features or attributes (reducts) from a large dataset that effectively predict the risk of diabetes at an early stage. The architecture of the hardware accelerator is optimized to efficiently handle large datasets commonly encountered in medical applications. It includes modules for data preprocessing, discernibility, matrix computation, and reduct determination. The proposed design is implemented on SPARTAN -6 Field Programmable Gate Array; it works on 100 MHz and supports 3 stages pipelining. The extensive experimentation, employing a comprehensive dataset of diabetes patients, unequivocally demonstrates the hardware accelerator's superior performance compared to traditional methods and software-based implementations. The Exact Reduct Rough Set Hardware Accelerator provides a powerful tool for medical professionals to identify patients at risk early on, enabling timely interventions and personalized care to improve patient health outcomes. The proposed accelerator contributes to the growing field of medical decision support systems and holds promise for broader applications in healthcare and beyond. The findings from this study can be seamlessly incorporated into the system-on-chip platform, contributing to the creation of intelligent tools for monitoring and managing diabetes. By offloading rough set computations to hardware accelerators on the wearable device, the data does not need to be continuously transmitted to external servers for analysis, ensuring user privacy and reducing communication overhead.

On-Device Execution of Deep Learning Models on HoloLens2 for Real-Time Augmented Reality Medical Applications.

The integration of Deep Learning (DL) models with the HoloLens2 Augmented Reality (AR) headset has enormous potential for real-time AR medical applications. Currently, most applications execute the models on an external server that communicates with the headset via Wi-Fi. This client-server architecture introduces undesirable delays and lacks reliability for real-time applications. However, due to HoloLens2's limited computation capabilities, running the DL model directly on the device and achieving real-time performances is not trivial. Therefore, this study has two primary objectives: (i) to systematically evaluate two popular frameworks to execute DL models on HoloLens2-Unity Barracuda and Windows Machine Learning (WinML)-using the inference time as the primary evaluation metric; (ii) to provide benchmark values for state-of-the-art DL models that can be integrated in different medical applications (e.g., Yolo and Unet models). In this study, we executed DL models with various complexities and analyzed inference times ranging from a few milliseconds to seconds. Our results show that Unity Barracuda is significantly faster than WinML (p-value < 0.005). With our findings, we sought to provide practical guidance and reference values for future studies aiming to develop single, portable AR systems for real-time medical assistance.

Failure Mode and Effects Analysis Prior to the Introduction of AI Generated GTVs for Brain Metastases in the Clinical Workflow

AI autosegmentation of organs-at-risk (OARs) is common practice at many radiotherapy clinics. Despite the abundance of gross tumor volume (GTV) autosegmentation algorithms, adoption in clinical care has been slow due to the high risk associated with errors in GTV delineation. Here we present a failure mode and effects analysis (FMEA) to evaluate the risk associated with introducing AI derived GTVs in patients treated with stereotactic radiosurgery (SRS). An AI GTV autosegmentation algorithm for brain metastases was developed in-house based on a V-Net 3D CNN. Registered CT and MR images and a contour of the brain are input into the software and all identified lesions are returned in a DICOM-RT structure set. Following algorithm evaluation, a workflow was developed to enable AI GTV autosegmentation to be introduced clinically for every SRS patient. The following steps were added to existing procedures: 1) workflow to send CT/MR and brain structure to external server, 2) autosegmentation run on the server, 3) AI GTV structures with a standard nomenclature added to existing OAR structure set, and 4) MD review, editing, and approval of AI GTVs. After successfully completing the physics evaluation testing of the new process, we formed a team of 10 faculty and staff including physicists, residents, physicians, and planners to perform the FMEA prior to clinical implementation. The team met to map the process, identify potential failure modes, and score their frequency of occurrence, severity, and detectability. A 3-point scale (1, 3, or 5) was used to simplify the scoring process. Occurrence was defined as rare, sometimes, or often; severity as low, medium, or high; and detectability as obvious, possible, or challenging. The risk probability numbers (RPNs) were calculated and the steps in the process with the highest RPNs were flagged for further discussion. The FMEA team completed their process map and analysis primarily in 4 meetings. The process map began with acquisition of the patients CT simulation scan and ended with physician approval of final volumes for treatment planning. We identified 17 process steps and 72 possible failure modes, of which 26 were associated with the new workflow. Eighteen failure modes had an RPN greater than 30 (highest risk score in at least one category) and were flagged to assess mitigation strategies. Five were unique to the new AI GTV workflow and mitigation strategies will be designed prior to clinical use. Those involved risks related to inaccurate AI GTV contours, false positives, and an incomplete review stemming from over-reliance by team members on AI. AI is increasingly being employed at every step of radiotherapy to automate and streamline processes. The FMEA analysis resulted in the identification of the riskiest parts of using AI GTV autosegmentation. This can be an effective tool in the development of checks to ensure that GTV autosegmentation methods can be safely introduced in support of patient care.

Accuracy of the evaluation of forest areas based on Landsat data using free software

Abstract Ever-evolving technologies are enabling us to obtain information about the world around us ever more quickly and precisely. This state of affairs contributes to the growing need to store and analyse data. For today’s scientists, this is a challenge because it involves analyses on a global scale. This also applies to spatial data, vast amounts of which are made available online. The Google Earth Engine platform is such a place on the web. It is not just a catalogue for browsing, but above all an environment for programming useful applications. Among free software, it is difficult to find one that is not dependent on the parameters of the computer. In the case of Google Earth Engine, the processes programmed by the user are executed on powerful external servers, and the user only gets the finished result, which he can download to his computer and use in further work. The initial chapters introduce the basic concepts and characterise the specifics of working in this environment, taking into account the limitations of the platform. Then, the individual stages of the algorithm developed by the authors are described, trying to explain well the reasons for the use of particular methods and functions. In order to verify the correctness of the obtained results, existing databases on the subject and published research results of other authors were used.

Self-x based closed loop wearable IoT for real-time detection and resolution of sleep apnea

This study introduces the Sleepcare Kit and S-Pillow (SKP), a wearable body-area network concept designed to address the uncomfortable and imprecise measurements associated with previous sleep apnea diagnosis and treatment methods that depend on external servers or human experts. SKP measures respiration, acceleration, and SpO2 signals to accurately identify risky sleep apnea in real-time. In addition, it incorporates a S-Pillow that can deterministically adjust the position of the patient’s head to resolve sleep apnea symptoms. SKP operates in a self-x based closed loop to optimally manage sleep apnea risk in a patient’s home environment without external help. Experimental results shows a 1/6 reduction in control frequency and a 1/3 reduction in apnea time compared to previous studies related to pillow use.

A Privacy-Preserving Blockchain Platform for a Data Marketplace

Recent data leak scandals, together with the under-utilization of collected data (estimated that around 90% of data never leaves a device’s local storage), limits the applicability and potential of novel data driven applications. Thus, novel ways to treat data, in which users are guaranteed control, usability, and privacy over their generated data are needed. In this paper we propose a novel privacy-preserving blockchain framework for a data sharing marketplace. The proposed framework allows users (such as sensors and devices) that generate data to store it in external servers, while the blockchain is utilized to record buy and sell transactions between parties, as well as perform access control by generating access sequences whenever trades are performed. A novel perspective over data ownership is presented, in which whoever generates the data has completed ownership and control over it and the blockchain transactions are only utilized to guarantee temporary access to it. The proposed blockchain framework also supports different types of data and provides, via the distributed and openness of the framework, quality, timeliness and similarity control over the data stored in the marketplace. In this context, different types of applications that can benefit from this framework are presented and open problems are discussed.