Digital Health Bonn

@article{gress2026c,

title = {BLE Security Testing: Survey of Attacks and Evaluation of Tools and Frameworks},

author = {Hannah Greß and Jonas Schreiter and Michael Rademacher and Björn Krüger },

doi = {10.1109/OJCOMS.2026.3691569},

year  = {2026},

date = {2026-05-08},

urldate = {2026-05-08},

journal = {IEEE Open Journal of the Communications Society},

abstract = {Nowadays, more and more areas of our lives are becoming ‘smart,’ including homes, industrial

sites, and medical devices. Since sensitive data is transmitted over various protocols, manufacturers must

ensure that the data is adequately secured. To do so, penetration testing on these devices is an option,

ideally before market launch. Depending on the protocol, various tools and frameworks exist. We chose

the Bluetooth Low Energy (BLE) protocol for analysis due to its widespread use and started our work by

classifying possible threats to BLE into four categories based on the attacks executable by the identified

tools and frameworks. Threats targeting BLE Mesh and BLE privacy and localization were identified

through a literature review and added to the corresponding categories of our taxonomy. Subsequently, we

conducted an in-depth evaluation of these tools and frameworks targeting BLE (excluding BLE Mesh,

and BLE privacy and localization) on four medical devices to empirically determine which tools and

frameworks remain usable for pentesting BLE devices. Furthermore, we analyzed which threats in our

taxonomy remain feasible. Our results show that only eight out of 21 (38%) attacks can still be carried

out. The tools and frameworks capable of conducting these attacks are mostly still maintained, emphasizing

the importance of keeping such tools and frameworks up to date for future use. Nevertheless, current BLE

pentesting requires a patchwork of multiple, often unmaintained frameworks to realize all known attacks

— this demonstrates an urgent need for a single, extensible, actively maintained framework that allows

attacks to be integrated modularly.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{Imran2026a,

title = {From Steps to Sentiments: Cross-Domain Transfer Learning for Activity-Based Emotion Detection in Wearable IoT Systems},

author = {Hamza Ali Imran and Qaiser Riaz and Kiran Hamza and Shaida Muhammad and Björn Krüger},

doi = {10.1109/JIOT.2026.3666469},

year  = {2026},

date = {2026-02-20},

urldate = {2026-02-20},

journal = {IEEE Internet of Things Journal},

pages = {1-1},

abstract = {Context-aware, gait-based sentiment analysis and emotion perception is an emerging research area within Internet of Things (IoT), aiming to make smart systems more intuitive and responsive. Recognizing emotions from wearable inertial sensor data is challenging due to subtle and compound emotional cues, variability across individuals and contexts, and limited, imbalanced datasets. To address these challenges, we propose Jazbat-Net, a lightweight neural network that leverages Transfer Learning (TL). The model is first trained on a large-scale, publicly available multi-activity dataset collected using wearable inertial sensors, and then retrained on a multi-class emotion dataset, effectively transferring knowledge from the pretraining phase. We evaluate Jazbat-Net with and without TL, across both smartwatch and smartphone based data, and for input dimensions ranging from 1D to 6D. The best results are achieved when pretrained on smartphone-based activity data and retrained on smartphone-based emotion data using a 1D input size. The proposed model attains an average classification accuracy of 95%, with a precision score of 95%, a recall score of 97%, and an F1-score of 96%. Moreover, Jazbat-Net achieves a low theoretical time complexity and requires only ≈ 6.96 M Multiply–Accumulate Operations (MACs), which is about 95% fewer computations than the previous State-of-the-Art (SOTA) model. Its space complexity is also low, with a model size of only ≈ 110 KB and peak activation memory of ≈ 0.35 MB. On-device evaluation on a Xiaomi 13T smartphone demonstrates that Jazbat-Net achieves a median inference latency of only ≈ 90.96 ms with a TFLite 32-bit floating point precision (FP32) model size of just ≈ 0.158 MB, making it ≈ 20× smaller and ≈ 20% faster than the previous SOTA model while maintaining comparable accuracy.

},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{Goharinejad-2025,

title = {Assessing the Impact of Virtual Reality, Augmented Reality, and Video Games on Improving Post-Traumatic Stress Disorder Symptoms: A Systematic Review and Meta-Analysis},

author = {Saeideh Goharinejad and Salime Goharinezhad and Khadijeh Moulaei and Björn Krüger and Thomas Spittler},

url = {https://doi.org/10.1177/00469580251413101},

doi = {10.1177/00469580251413101},

year  = {2026},

date = {2026-01-28},

urldate = {2025-12-01},

journal = {INQUIRY: The Journal of Health Care Organization, Provision, and Financing},

volume = {63},

pages = {00469580251413101},

abstract = {Post-traumatic stress disorder (PTSD) is often debilitating, with current treatments limited by low adherence, high costs, and accessibility issues. Innovative technologies such as virtual reality (VR), augmented reality (AR), and therapeutic video games provide immersive environments that may improve treatment outcomes. This systematic review and meta-analysis evaluated the efficacy of these approaches and explored their potential advantages over traditional methods. A comprehensive search of PubMed, PsycINFO, CINAHL, Web of Science, and Cochrane identified relevant studies. Two reviewers independently screened articles, extracted data, and assessed quality using the Mixed Methods Appraisal Tool (MMAT). A random-effects model was used to calculate pooled effect sizes (Hedges’ g), and heterogeneity was evaluated with the Q test and I2 statistic. Publication bias was examined with funnel plots, Egger’s, and Begg’s tests. Analyses were performed in Stata version 17.0. From 480 records, 21 studies were included in the review and 12 in the meta-analysis. VR-based treatments yielded a pooled effect size of –0.35 (95% CI [–0.57, –0.13]), indicating a small-to-moderate reduction in PTSD symptoms. The effect was statistically significant (z = –3.13, P < .01), with moderate heterogeneity (I2 = 46.28%, P = .03). Funnel plots and statistical tests suggested minimal publication bias. Meta-regression showed no moderating effect of gender. Subgroup analyses indicated significant benefits in male-only samples, participants aged 20 to 30 and over 40, and studies with follow-up periods ≤7 months. Larger effects were observed in studies with 15 to 30 participants. VR, AR, and video game interventions significantly reduce PTSD symptoms and may enhance accessibility and engagement compared to traditional treatments. These findings support the integration of immersive technologies into therapeutic practice to improve outcomes for individuals with PTSD. }

},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{Bhatti2025,

title = {Beyond Falls: A Hybrid CNN–LSTM–Attention Framework for Pre-, Transition-, and Post-Fall Detection with Wearable Inertial Sensors},

author = {Faraz Ahmad Bhatti and Qaiser Riaz and Björn Krüger},

doi = {10.1109/ACCESS.2025.3641198},

year  = {2025},

date = {2025-12-05},

urldate = {2025-12-02},

journal = {IEEE Access},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{Moontaha2025,

title = {EPIStress: A multimodal dataset of Physiological signals to measure cognitive stress in epilepsy patients},

author = {Sidratul Moontaha and Constanze Cavalier and Birgitta Esser and Arthur Jordan and Ines Goebel and Christoph Anders and Afsana Mimi and Björn Krüger and Rainer Surges and Bert Arnrich},

url = {https://doi.org/10.1038/s41597-025-06328-3},

doi = {10.1038/s41597-025-06328-3},

isbn = {2052-4463},

year  = {2025},

date = {2025-11-28},

urldate = {2025-12-01},

journal = {Scientific Data},

volume = {12},

number = {1867},

issue = {1},

abstract = {Epilepsy patients commonly report stress as a frequent seizure trigger; however, the objective seizure-stress relationship is unclear due to self-report biases and difficulty in objective quantification of stress. This work presents a dataset from twenty epilepsy patients undergoing cognitive stress elicitation protocols, participating in laboratory experiments with computer-based tasks at predefined difficulty levels, and in situational experiments by independently choosing tasks with at least two difficulty levels. Physiological signals from wearable electroencephalography, photoplethysmography, acceleration, electrodermal activity, and temperature sensors were recorded. The task-related perceived cognitive stress was collected using two 5-point Likert scales of self-reported mental workload and stress, contrasted by a pairwise NASA-TLX questionnaire. Additionally, the dataset includes a patient-reported list of seizure-provoking and -inhibiting factors. Results illustrated individual and heterogeneous responses to cognitive tasks, with some modalities yielding statistically significant features, while others demonstrated expected directional trends. The findings support the validity and suitability of the proposed dataset for cognitive stress detection and the potential to map seizure-related factors to cognitive stress events.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{alhaj2025,

title = {Eyes on Cognition: Exploring Oculomotor Correlates of Cognitive Function in Patients with Epilepsy},

author = {Sarah Al-Haj Mustafa and Anna Jansen and Melissa Steininger and Johannes Müllers and Rainer Surges and Randi von Wrede and Björn Krüger and Christoph Helmstaedter},

doi = {10.1016/j.yebeh.2025.110562},

year  = {2025},

date = {2025-06-30},

urldate = {2025-06-30},

journal = {Epilepsy & Behavior},

volume = {173},

number = {110562},

issue = {December 2025},

abstract = {Objective

This study investigates the relationship between eye tracking parameters and cognitive performance during the Trail Making Test (TMT) in individuals with epilepsy and healthy controls. By analyzing ocular behaviors such as saccade velocity, fixation duration, and pupil diameter, we aim to determine how these metrics reflect executive functioning and attentional control.

Methods

A sample of 95 participants with epilepsy and 34 healthy controls completed the TMT while their eye movements were recorded. Partial correlations, controlling for age, sex, education, medication count, seizure status and epilepsy duration, examined associations between eye tracking measures and cognitive performance derived from EpiTrack and TMT performance.

Results

In the patient group, faster TMT-A performance was associated with shorter fix- ation durations (r = 0.31, p = 0.006). Lower minimum saccade velocity correlated with slower performance on both TMT-A (r = −0.35, p = 0.002) and TMT-B (r = −0.40, p<0.001), whereas higher peak saccade velocities were linked to worse performance (TMT-A: r = 0.45, p<0.001; TMT-B: r = 0.41, p<0.001). Pupil diameter findings indicated that slower TMT performance was associated with smaller minimum pupil sizes (r = −0.23 to r = −0.36), wich may indicate increased cognitive effort and attentional load. Higher EpiTrack scores also correlated with a smaller minimum pupil diameter − but only during the more demanding TMT-B − and with a more restricted saccade velocity range, reflecting greater motor control and attentional stability. No significant correlations emerged within the control group.

Conclusion

These findings highlight the potential of eye tracking as a non-invasive tool for assessing cognitive function in epilepsy. Efficient cognitive performance was characterized by stable and controlled eye movements, whereas impaired performance involved erratic saccade dynamics and prolonged fixations. Importantly, eye tracking parameters provide additional information beyond simple speed measurements, potentially enhancing the differential diagnostic capabilities of the TMT in epilepsy. The observed associations between oculomotor parameters and cognitive performance were not present in the control group, suggesting that these relationships may be specific to epilepsy. Future research should investigate whether both basic and advanced metrics of search strategies are sensitive to disease dynamics and treatment effects in epilepsy.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{2025-khan,

title = {Towards Effective Parkinson’s Monitoring: Movement Disorder Detection and Symptom Identification Using Wearable Inertial Sensors},

author = {Umar Khan and Qaiser Riaz and Mehdi Hussain and Muhammad Zeeshan and Björn Krüger},

url = {https://www.mdpi.com/1999-4893/18/4/203},

doi = {10.3390/a18040203},

issn = {1999-4893},

year  = {2025},

date = {2025-04-04},

urldate = {2025-01-01},

journal = {Algorithms},

volume = {18},

number = {4},

abstract = {Parkinson’s disease lacks a cure, yet symptomatic relief can be achieved through various treatments. This study dives into the critical aspect of anomalous event detection in the activities of daily living of patients with Parkinson’s disease and the identification of associated movement disorders, such as tremors, dyskinesia, and bradykinesia. Utilizing the inertial data acquired from the most affected upper limb of the patients, this study aims to create an optimal pipeline for Parkinson’s patient monitoring. This study proposes a two-stage movement disorder detection and classification pipeline for binary classification (normal or anomalous event) and multi-label classification (tremors, dyskinesia, and bradykinesia), respectively. The proposed pipeline employs and evaluates manual feature crafting for classical machine learning algorithms, as well as an RNN-CNN-inspired deep learning model that does not require manual feature crafting. This study also explore three different window sizes for signal segmentation and two different auto-segment labeling approaches for precise and correct labeling of the continuous signal. The performance of the proposed model is validated on a publicly available inertial dataset. Comparisons with existing works reveal the novelty of our approach, covering multiple anomalies (tremors, dyskinesia, and bradykinesia) and achieving 93.03% recall for movement disorder detection (binary) and 91.54% recall for movement disorder classification (multi-label). We believe that the proposed approach will advance the field towards more effective and comprehensive solutions for Parkinson’s detection and symptom classification.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{2025gressBT,

title = {The Newer, the More Secure? Standards-Compliant Bluetooth Low Energy Man-in-the-Middle Attacks on Fitness Trackers},

author = {Hannah Greß and Björn Krüger and Elmar Tischhauser},

url = {https://www.mdpi.com/1424-8220/25/6/1815},

doi = {10.3390/s25061815},

issn = {1424-8220},

year  = {2025},

date = {2025-03-14},

urldate = {2025-01-01},

journal = {Sensors},

volume = {25},

number = {6},

abstract = {The trend in self-tracking devices has remained unabated for years. Even if they record a large quantity of sensitive data, most users are not concerned about their data being transmitted and stored in a secure way from the device via the companion app to the vendor’s server. However, the secure implementation of this chain from the manufacturer is not always given, as various publications have already shown. Therefore, we first provide an overview of attack vectors within the ecosystem of self-tracking devices. Second, we evaluate the data security of eight contemporary fitness trackers from leading vendors by applying four still partly standards-compliant Bluetooth Low-Energy Man-in-the-Middle (MitM) attacks. Our results show that the examined devices are partially vulnerable against the attacks. For most of the trackers, the manufacturers put different security measures in place. These include short and user-initiated visibility and connectivity or app-level authentication to limit the attack surface. Interestingly, newer models are more likely to be attackable, underlining the constant need for verifying the security of BLE devices, reporting found vulnerabilities, and also strengthening standards and improving security awareness among manufacturers and users. Therefore, we finish our work with recommendations and best practices for law- and regulation-makers, vendors, and users on how to strengthen the security of BLE devices.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{10552639,

title = {Unveiling Fall Origins: Leveraging Wearable Sensors to Detect Pre-Impact Fall Causes},

author = {Samia Kiran and Qaiser Riaz and Mehdi Hussain and Muhammad Zeeshan and Björn Krüger},

doi = {10.1109/JSEN.2024.3407835},

issn = {1558-1748},

year  = {2024},

date = {2024-08-01},

urldate = {2024-01-01},

journal = {IEEE Sensors Journal},

volume = {24},

number = {15},

pages = {24086-24095},

abstract = {Falling poses a significant challenge to the health and well-being of the elderly and people with various disabilities. Precise and prompt fall detection plays a crucial role in preventing falls and mitigating the impact of injuries. In this research, we propose a deep classifier for pre-impact fall detection which can detect a fall in the pre-impact phase with an inference time of 46–52 milliseconds. The proposed classifier is an ensemble of Convolutional Neural Networks (CNNs) and Bidirectional Gated Recurrent Units (BiGRU) with residual connections. We validated the performance of the proposed classifier on a comprehensive, publicly available preimpact fall dataset. The dataset covers 36 diverse activities, including 15 types of fall-related activities and 21 types of activities of daily living (ADLs). Furthermore, we evaluated the proposed model using three different inputs of varying dimensions: 6D input (comprising 3D accelerations and 3D angular velocities), 3D input (3D accelerations), and 1D input (magnitude of 3D accelerations). The reduction in the input space from 6D to 1D is aimed at minimizing the computation cost. We have attained commendable results outperforming the state-of-the-art approaches by achieving an average accuracy and F1 score of 98% for 6D input size. The potential implications of this research are particularly relevant in the realm of smart healthcare, with a focus on the elderly and differently-abled population.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{s24196167,

title = {Generisch-Net: A Generic Deep Model for Analyzing Human Motion with Wearable Sensors in the Internet of Health Things},

author = {Kiran Hamza and Qaiser Riaz and Hamza Ali Imran and Mehdi Hussain and Björn Krüger},

url = {https://www.mdpi.com/1424-8220/24/19/6167},

doi = {10.3390/s24196167},

issn = {1424-8220},

year  = {2024},

date = {2024-01-01},

urldate = {2024-01-01},

journal = {Sensors},

volume = {24},

number = {19},

abstract = {The Internet of Health Things (IoHT) is a broader version of the Internet of Things. The main goal is to intervene autonomously from geographically diverse regions and provide low-cost preventative or active healthcare treatments. Smart wearable IMUs for human motion analysis have proven to provide valuable insights into a person’s psychological state, activities of daily living, identification/re-identification through gait signatures, etc. The existing literature, however, focuses on specificity i.e., problem-specific deep models. This work presents a generic BiGRU-CNN deep model that can predict the emotional state of a person, classify the activities of daily living, and re-identify a person in a closed-loop scenario. For training and validation, we have employed publicly available and closed-access datasets. The data were collected with wearable inertial measurement units mounted non-invasively on the bodies of the subjects. Our findings demonstrate that the generic model achieves an impressive accuracy of 96.97% in classifying activities of daily living. Additionally, it re-identifies individuals in closed-loop scenarios with an accuracy of 93.71% and estimates emotional states with an accuracy of 78.20%. This study represents a significant effort towards developing a versatile deep-learning model for human motion analysis using wearable IMUs, demonstrating promising results across multiple applications.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{s23073664,

title = {An Effective and Efficient Approach for 3D Recovery of Human Motion Capture Data},

author = {Hashim Yasin and Saba Ghani and Björn Krüger},

url = {https://www.mdpi.com/1424-8220/23/7/3664},

doi = {10.3390/s23073664},

issn = {1424-8220},

year  = {2023},

date = {2023-01-01},

journal = {Sensors},

volume = {23},

number = {7},

abstract = {In this work, we propose a novel data-driven approach to recover missing or corrupted motion capture data, either in the form of 3D skeleton joints or 3D marker trajectories. We construct a knowledge-base that contains prior existing knowledge, which helps us to make it possible to infer missing or corrupted information of the motion capture data. We then build a kd-tree in parallel fashion on the GPU for fast search and retrieval of this already available knowledge in the form of nearest neighbors from the knowledge-base efficiently. We exploit the concept of histograms to organize the data and use an off-the-shelf radix sort algorithm to sort the keys within a single processor of GPU. We query the motion missing joints or markers, and as a result, we fetch a fixed number of nearest neighbors for the given input query motion. We employ an objective function with multiple error terms that substantially recover 3D joints or marker trajectories in parallel on the GPU. We perform comprehensive experiments to evaluate our approach quantitatively and qualitatively on publicly available motion capture datasets, namely CMU and HDM05. From the results, it is observed that the recovery of boxing, jumptwist, run, martial arts, salsa, and acrobatic motion sequences works best, while the recovery of motion sequences of kicking and jumping results in slightly larger errors. However, on average, our approach executes outstanding results. Generally, our approach outperforms all the competing state-of-the-art methods in the most test cases with different action sequences and executes reliable results with minimal errors and without any user interaction.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{yasin-2021a,

title = {An Efficient 3D Human Pose Retrieval and Reconstruction from 2D Image-Based Landmarks},

author = {Hashim Yasin and Björn Krüger},

url = {https://www.mdpi.com/1424-8220/21/7/2415

https://digital-health-bonn.de/wp-content/uploads/2024/03/sensors-21-02415.pdf, Paper},

doi = {10.3390/s21072415},

issn = {1424-8220},

year  = {2021},

date = {2021-01-01},

urldate = {2021-01-01},

journal = {Sensors},

volume = {21},

number = {7},

abstract = {We propose an efficient and novel architecture for 3D articulated human pose retrieval and reconstruction from 2D landmarks extracted from a 2D synthetic image, an annotated 2D image, an in-the-wild real RGB image or even a hand-drawn sketch. Given 2D joint positions in a single image, we devise a data-driven framework to infer the corresponding 3D human pose. To this end, we first normalize 3D human poses from Motion Capture (MoCap) dataset by eliminating translation, orientation, and the skeleton size discrepancies from the poses and then build a knowledge-base by projecting a subset of joints of the normalized 3D poses onto 2D image-planes by fully exploiting a variety of virtual cameras. With this approach, we not only transform 3D pose space to the normalized 2D pose space but also resolve the 2D-3D cross-domain retrieval task efficiently. The proposed architecture searches for poses from a MoCap dataset that are near to a given 2D query pose in a definite feature space made up of specific joint sets. These retrieved poses are then used to construct a weak perspective camera and a final 3D posture under the camera model that minimizes the reconstruction error. To estimate unknown camera parameters, we introduce a nonlinear, two-fold method. We exploit the retrieved similar poses and the viewing directions at which the MoCap dataset was sampled to minimize the projection error. Finally, we evaluate our approach thoroughly on a large number of heterogeneous 2D examples generated synthetically, 2D images with ground-truth, a variety of real in-the-wild internet images, and a proof of concept using 2D hand-drawn sketches of human poses. We conduct a pool of experiments to perform a quantitative study on PARSE dataset. We also show that the proposed system yields competitive, convincing results in comparison to other state-of-the-art methods.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}