Privacy-Preserving Data Aggregation Techniques for Wireless Body Area Networks: A Comprehensive Survey and Comparative Analysis

We suggest a classification that groups existing systems according to encryption, routing, context awareness, AI/ML integration, and hybrid approaches in order to offer a structured knowledge of the various privacy-preserving data aggregation strategies in WBANs. The hierarchical structure of these methods is depicted in Figure 1.

Figure 1.

Demonstrate how the underlying technology and operating principles of privacy-preserving aggregation techniques can be used to categories them into major groupings

The selection process for data aggregation techniques is a critical first step, involving a thorough review and analysis of existing methods in the context of WBANs. Techniques such as Quantum Key Distribution (QKD), Attribute-Based Encryption (ABE), and Secure Multi-path Data Aggregation have been identified as promising candidates due to their potential to significantly enhance the privacy, security, and overall efficiency of WBANs. The selection criteria are grounded in an extensive review of the literature, focusing on factors like scalability, computational efficiency, security robustness, and compatibility with the resource-constrained nature of WBANs. By critically evaluating these techniques against the identified criteria, the study ensures that only the most suitable methods are chosen for further analysis and implementation.

Clinical datasets do not directly generate network traffic; therefore, they were transformed into WBAN-compatible workloads. Each dataset was processed to create realistic sensor data streams. Clinical datasets such as MIMIC-III and PhysioNet were used only as Payload generators to emulate realistic health-monitoring traffic in the NS-3 environment. Each patient record was mapped to a set of virtual body sensors (ECG, SpO2, temperature, and motion). For simulation, feature including heart rate, oxygen saturation, and temperature were encoded as 32-byte payloads transmitted once per second (1 Hz sampling rate). Each packet therefore contained 128 bytes of sensor payload plus 16 bytes of protocol headers. All payloads were numerically normalized and anonymized prior to transmission. Thus, the datasets served to model data volume and timing not to evaluate clinical accuracy. The dataset usage details are summarized in Table 2.

The data used in this study are sourced from both publicly available datasets and simulated environments to ensure a comprehensive evaluation. Public sources such as the MIMIC-III and PhysioNet databases provide real-world patient data that is crucial for testing the effectiveness of the data aggregation techniques in realistic scenarios. Additionally, simulated data is generated to model specific conditions that may not be fully captured in public datasets. Prior to analysis, the data undergoes a rigorous preprocessing stage, which includes normalization. Before comparative analysis, data values are subject to a standard min-max normalization procedure in order to ensure consistency across heterogenous data sources and performance metrics. The normalization process, defined by the equation (1) 1Xnorm=X−XminXmax−Xmin{X_{{\rm{norm }}}} = {{X - {X_{\min }}} \over {{X_{\max }} - {X_{\min }}}}

Normalization is applied on a per-feature basis, allowing values with different units and ranges to be represented on a common scale. where X_norm represents the corresponding normalized value, the original data for a given feature, and X_min, X_max represent the minimum and maximum values of that feature within the considered dataset.

To model successful packet reception in on-body links we adopt a distance dependent probabilistic reception model that reflects the combined effects of path loss and small-scale fading in short-range WBAN links. The reception probability for a packet transmitted by source node S and received (potentially) by multiple candidate receivers is modelled as (equation (2)): 2P(s→i)=exp(−λdsi)∑j=inexp(−λdsj)P(s \to i) = {{\exp \left( { - \lambda {d_{si}}} \right)} \over {\sum\nolimits_{j = i}^n {exp} \left( { - \lambda {d_{sj}}} \right)}}

Where P(s → i) is the normalized probability that receiver I successfully receives the packet from source S (conditional on there being at least on reception attempt).

d_si is the Euclidean distance between source S and receiver i.

λ is an attenuation parameter (path-loss exponent surrogate) that controls the rate of exponential decay with distance, and

The denominator ∑j=inexp(−λdsj)\sum\nolimits_{j = i}^n {exp} \left( { - \lambda {d_{sj}}} \right) is a normalization constant that ensures ∑i=1np(s→i)=1\sum\nolimits_{i = 1}^n p (s \to i) = 1 across the set of candidate receivers.

Each technique is implemented within the simulation framework. For instance, the QKD algorithm is modelled to evaluate its key generation and distribution efficiency under varying network conditions. The key rate, R, is calculated using the equation (3): 3R=q·IAB−f·q·H(E)R = q\cdot{I_{AB}} - f\cdotq\cdotH(E) where q is the sifting factor, I_AB is the mutual information between Alice and Bob, f is the error correction efficiency, and H(E) is the binary entropy function of the quantum bit error rate E.

The parameter reporter in this Table 3 (e.g., key length, encryption type, or security level) represents typical configurations commonly cited in prior WBAN literature and are include for comparative reference only.

The performance of each data aggregation technique is systematically tested and evaluated against a set of predefined metrics. These metrics include energy consumption, efficiency, reliability, and security robustness. The energy consumption E_c for each technique is evaluated using equation (4): 4Ec=∫p(t)dt{E_c} = \int p (t)dt where P (t) represents the power consumption over time t. Efficiency and reliability metrics are similarly calculated, providing a comprehensive view of each technique’s performance.

After the testing phase, the collected performance data is subjected to detailed statistical analysis to interpret the results and draw meaningful conclusions. Techniques such as ANOVA (Analysis of Variance) are employed to discern significant differences in the performance of the data aggregation techniques. This statistical approach helps in identifying which techniques are most effective under specific conditions and for particular WBAN applications.

The results are then analyzed to provide insights into the optimal data aggregation approaches, offering recommendations for enhancing the privacy, security, and efficiency of WBANs. The interpretation of these results is crucial for guiding future research and development in the field of WBANs, ensuring that the most promising techniques are further explored and refined.

Multiple datasets are used to validate and evaluate privacy-preserving data aggregation methods in Wireless Body Area Networks (WBANs), each one tailored to specific experimental requirements. The MIMIC-III dataset from Beth Israel Deaconess Medical Center includes comprehensive patient vitals, lab tests, and survival data from approximately 40,000 patients, ideal for intensive care health monitoring scenarios. PhysioNet offers a variety of physiological signals across different datasets, facilitating the real-time testing of algorithmic performance. The Arrhythmia Dataset and the Heart Disease Dataset, both hosted by the UCI Machine Learning Repository, contain specific clinical records for cardiac condition analysis and disease prediction respectively, with the former providing 452 instances and the latter 303 instances. For broader activity-based health tracking, the Healthcare Activity Monitoring dataset available on Kaggle includes data on activity, heart rate, and sleep patterns from around 10,000 instances. Lastly, custom simulations of WBAN data provide configurable datasets tailored to specific research scenarios, allowing researchers to test under various hypothetical yet realistic WBAN conditions. Each dataset is freely accessible, though MIMIC-III requires approval due to the sensitivity of the information, ensuring that researchers can obtain and use the data under ethical guidelines.

The datasets mentioned are instrumental in advancing the research on privacy-preserving data aggregation techniques within Wireless Body Area Networks, due to their diverse and rich data characteristics. The MIMIC-III dataset, sourced from a clinical setting, provides a robust foundation for analyzing complex medical scenarios, particularly in intensive care, where high fidelity data such as patient vitals and lab tests are critical for testing the efficacy and reliability of new data aggregation methods. PhysioNet extends the scope by offering a varied collection of physiological signals that are essential for validating real-time signal processing algorithms under diverse health monitoring situations. The Arrhythmia and Heart Disease Datasets from the UCI Machine Learning Repository are smaller in scale but highly focused, allowing for detailed studies on specific cardiovascular conditions, thereby helping in refining algorithms aimed at cardiological data aggregation. On a broader scale, the Healthcare Activity Monitoring dataset from Kaggle encompasses data on general physical activities and biometric monitoring, useful for applications in lifestyle and fitness tracking in WBANs. Finally, the Simulated WBAN Data offers unparalleled flexibility by allowing researchers to model and manipulate data according to custom specifications and scenarios, thus facilitating the development of tailored solutions that can be benchmarked against real-world data conditions before actual deployment. Each dataset’s accessibility and specificity provide unique opportunities for addressing distinct aspects of privacy and efficiency in data aggregation within WBAN environments. Table 4. Representative datasets frequently referenced in existing literature for evaluating data aggregation and privacy mechanisms in Wireless Body Area Networks (WBANs). This survey does not implement or execute cryptographic operations within the NS-3 simulation environment. Instead, network performance indicators (latency, energy consumption, packet deliverytrends) are derived from analytical and modelling assumptions reported in the corresponding source studies. Therefore, the listed values should be interpreted as representative characteristics rather than experimentally validated measurements under a unified simulation setup.

To ensure transparency and repeatability, privacy-preservation and data-integrity scores were assigned based on a standardized rubric. Each technique was evaluated against five measurable criteria, with a total score computed as the sum of criterion weights. The privacy preservation scoring criteria are presented in Table 5, and the data integrity scoring criteria are presented in Table 6.

In a comprehensive evaluation of data aggregation techniques applicable to Wireless Body Area Networks (WBANs), major advanced methods were analysed for their efficacy across multiple performance indicators: privacy preservation, data integrity, throughput, latency, energy consumption, and accuracy of data aggregation. The Edge Computing-Based Aggregation technique demonstrates robust performance with a privacy preservation and data integrity score of 4 each, a high throughput of 340 data per minute, and low latency at 85 ms, making it highly suitable for real-time health monitoring that demands quick data processing and considerable privacy. Similarly, Quantum Key Distribution (QKD) and Lightweight Homomorphic Encryption both offer superior privacy and data integrity, with scores of 5, indicating their strong capability in securing data against unauthorized access and tampering. However, these techniques also exhibit higher latencies of 130 ms and 150 ms respectively, and increased energy consumption (0.60 J and 0.65 J), which might be prohibitive in environments where energy efficiency and speed are critical.

On the other hand, AI-Enhanced Anonymization and Differential Privacy Techniques, scoring slightly lower on data integrity, still maintain commendable privacy protection with scores of 4 and 5 respectively. They provide reasonable throughput and moderate latency, which balances the tradeoffs between operational speed and security. In particular, AI-Enhanced Anonymization achieves a throughput of 320 data/min and a latency of 95 ms, alongside a moderate energy consumption of 0.42 J, positioning it as a viable option for applications requiring a balance between data privacy and system responsiveness. Adaptive Streaming Filters, although the least secure with privacy and integrity scores of 3, excel in operational metrics, offering the highest throughput of 350 data/min and the lowest latency of 80 ms, coupled with the minimal energy demand of 0.33 J. This makes it an optimal choice for less sensitive applications where efficiency and low power consumption are paramount. The accuracy of data aggregation varies moderately among the techniques, with QKD leading at 95% accuracy, ensuring that the highest security does not compromise data usability. This detailed analysis illustrates in Figure 2 that while no single data aggregation technique universally outperforms in all categories, each offers distinct advantages and trade-offs, making them suitable for different applications within WBANs based on specific requirements for security, speed, and power efficiency. This nuanced understanding assists stakeholders in making informed decisions tailored to their particular operational needs and constraints. The performance values summarized in Table 7 reflect normalized or representative metrics extracted from prior studies. Scores assigned on a 1-5 scale for privacy presentation and data integrity represent relative assessments based on reported security capabilities and threat models, rather than empirically validated measurements obtained from a unified experimental framework.

Figure 2.

Visual comparison of key performance trends (Privacy, latency, throughput, and energy consumption) of WBAN data aggregation technique’s based on representation values reported in prior studies

Table 8 presents the network performance metrics for various data aggregation techniques in WBANs, highlighting differences in packet delivery ratio, error rate, network latency, and system efficiency. Network-level performance indicators play a critical role in determining the practical viability of data aggregation techniques in WBAN environments. Edge computing-based aggregation is frequently associated with high packet delivery ratios and low network delay due to localized processing and reduced communication overhead. QKD-based approaches, while highly secure procedure. Light weight Homomorphic Encryption introduces additional computational overhead, which may affect real-time performance in delay-sensitive scenarios.

In assessing the viability of various data aggregation techniques for WBANs requires careful consideration of network performance indicators that directly affect the effectiveness and dependability of data transfer in these networks. This paper presents a detailed examination of six data aggregation techniques, focusing on their packet delivery ratios, error rates, network latency, and overall system efficiency. Edge Computing-Based Aggregation demonstrates a strong performance with a packet delivery ratio of 98.5%, which is among the highest observed in our study. Coupled with a low error rate of 0.5% and an impressively short network latency of 85 ms, it achieves a system efficiency rating of 4. These metrics make it an ideal candidate for real-time health monitoring where timely and accurate data delivery is paramount. Quantum Key Distribution (QKD), known for its advanced security capabilities, shows a packet delivery ratio of 96.0%. Although slightly lower than some other techniques, it maintains an exceptionally low error rate of 0.2%. However, its higher latency of 130 ms and a system efficiency rating of 3 suggest some limitations in scenarios demanding rapid data transmission.

Lightweight Homomorphic Encryption presents a packet delivery ratio of 94.5%, the lowest among the techniques tested, and an error rate of 0.8%. With the highest latency recorded at 150 ms, its overall system efficiency is also rated at 3, indicating potential challenges in high-demand realtime operations. AI-Enhanced Anonymization and Differential Privacy Techniques both perform robustly with packet delivery ratios of 97.0% and 95.0%, respectively. AI-Enhanced Anonymization records an error rate of 0.6% and a latency of 95 ms, while Differential Privacy Techniques show a slightly higher error rate of 0.7% and latency of 105 ms. Both techniques are assigned a system efficiency rating of 4, balancing good security with respectable network performance. Lastly, Adaptive Streaming Filters stands out with the highest packet delivery ratio at 99.0% and an exceptionally low latency of 80 ms, alongside a minimal error rate of 0.4%. It is rated at 5 in system efficiency, making it exceptionally well-suited for applications where high throughput and low latency are critical.

The detailed metrics provided for each technique illustrate in Figure 3, the trade-offs between network efficiency, speed, and reliability in the context of WBANs. While some techniques offer higher security, they may sacrifice speed and efficiency, and vice versa. These insights allow researchers and practitioners to make informed decisions based on specific needs, such as the prioritization of realtime data processing or the uncompromised security of sensitive health data in WBAN environments. This nuanced approach to selecting data aggregation techniques based on precise network performance parameters ensures optimal deployment in targeted applications within the healthcare sector and beyond.

Figure 3.

Illustrative comparison of network performance trends among WBAN data aggregation techniques derived from reported packet delivery, latency, and efficiency metrics in existing research

Advanced cryptographic mechanisms such as homomorphic encryption and Quantum Key Distribution (QKD) offer strong privacy protection but are often associated with high computational and energy overhead. This trade-off poses a critical challenge for WBANs, where sensor nodes operate under strict energy constraints. Future research should explore lightweight security mechanisms, hardware-assisted cryptography, or hybrid approaches that balance security requirements with energy efficiency.

Many healthcare monitoring applications demand near real-time data transmission, where even minor delays can affect clinical responsiveness. However, several privacy-preserving aggregation techniques introduce latency due to complex encryption operations or multi-hop aggregation processes. Developing adaptive aggregation mechanisms that dynamically adjust security levels based on application urgency remains an open research problem.

As WBANs increasingly integrate heterogenous sensor devices and interface with broader IoTand hospital information systems, scalability becomes a significant concern. Existing aggregation schemes often assume homogenous node capabilities, limiting their applicability in real-world deployments. Future research should focus on scalable aggregation models that can efficiently support diverse node types and varying network sizes.

Patient mobility and body movements introduce frequent topology changes and fluctuating link quality in WBAN environments. Many current aggregation schemes rely on static network assumptions, which may not accurately reflect real-world conditions. Mobility-aware and context-adaptive aggregation techniques represent a critical research direction for improving system robustness.

While blockchain and Zero-Knowledge Proof (ZKP) mechanisms have demonstrated potential for privacy preservation in other domains, their application in WBANs remains limited due to resource constraints. Future studies may investigate lightweight blockchain architectures or simplified ZKP models tailored to the computational and energy limitations of WBAN systems.

This study did not involve any direct interaction with human participants. All clinical datasets used—MIMIC-III, PhysioNet, UCI, and Kaggle—are publicly available and fully de-identified. Access to the MIMIC-III (v1.4) database was obtained through the official PhysioNet Credentialed Access process. All authors who accessed MIMIC-III completed the required CITI Program “Data or Specimens Only Research” ethics certification, as mandated by the dataset’s usage policy. Because only de-identified secondary data were used, and no identifiable health information was accessed or redistributed, the study qualifies as IRB-exempt research under typical human-subjects’ regulations. No new identifiable patient data were collected.

To preserve privacy:

• 1.

  Raw clinical records were not used directly in any network evaluation.

• 2.

  All datasets were transformed into WBAN-compatible synthetic traffic streams (sampling rates, packet sizes, and payload patterns), ensuring no patient-level information entered the simulation.

• 3.

  No re-identification, linkage, or patient-level inference was performed.

• 4.

  No data were redistributed; all datasets were accessed and processed in accordance with their respective licenses.

MIMIC-III v1.4 is accessible via PhysioNet under credentialed access.

PhysioNet waveform datasets, UCI repositories, and Kaggle datasets are publicly available under their respective licenses.

All NS-3 simulation scripts and configuration files used to generate WBAN workloads will be released in an open repository upon publication (link to be activated in final version).

No proprietary or restricted data were generated in this study.