pith. sign in

arxiv: 2510.17162 · v2 · submitted 2025-10-20 · 💻 cs.LG

ALPINE: Closed-Loop Adaptive Privacy Budget Allocation for Mobile Edge Crowdsensing

Guanjie Cheng , Siyang Liu , Xinkui Zhao , Yishan Chen , Junqin Huang , Linghe Kong , Shiguang Deng This is my paper

Pith reviewed 2026-05-18 06:20 UTC · model grok-4.3

classification 💻 cs.LG
keywords mobile edge crowdsensingadaptive privacyprivacy budget allocationlocal differential privacyTD3 reinforcement learningrisk perceptionclosed-loop controledge computing
0
0 comments X p. Extension

The pith

ALPINE achieves a better privacy-utility trade-off in mobile edge crowdsensing through closed-loop adaptive privacy budget allocation using a TD3 policy.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes ALPINE as a lightweight closed-loop framework for adaptive privacy budget allocation in mobile edge crowdsensing. It models multi-dimensional risks including channel, semantic, contextual, and resource factors on terminal devices, then uses an offline-trained TD3 reinforcement learning policy to select appropriate privacy budgets. These budgets drive local differential privacy perturbations before data transmission, with edge-side evaluations providing feedback to refine the policy. This approach addresses the limitations of static or coarse-grained privacy schemes that struggle with dynamic conditions. If successful, it enables real-time privacy protection while maintaining data utility and low overhead on resource-constrained devices.

Core claim

ALPINE performs multi-dimensional risk perception on terminal devices by jointly modeling channel, semantic, contextual, and resource risks, and maps the resulting risk state to a privacy budget through an offline-trained TD3 policy. The selected budget drives local differential privacy perturbation, while edge-side privacy-utility evaluation provides feedback for policy switching and periodic refinement, forming a terminal-edge collaborative control loop for real-time, risk-adaptive privacy protection with low online overhead.

What carries the argument

The offline-trained TD3 policy that maps multi-dimensional risk states to privacy budgets within a terminal-edge feedback loop for adaptive allocation.

If this is right

  • ALPINE achieves a better privacy-utility trade-off than representative baselines on multiple real-world datasets.
  • It reduces the effectiveness of membership inference, property inference, and reconstruction attacks.
  • It preserves robust downstream task performance under dynamic risk conditions.
  • ALPINE introduces only modest runtime overhead on resource-constrained devices.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The framework could extend to other edge computing applications involving continuous data collection under varying risks.
  • Periodic policy refinement might allow adaptation to evolving threat landscapes without full retraining.
  • Testing the system in live deployments with diverse device types would validate its generalizability beyond the evaluated datasets.

Load-bearing premise

The offline-trained TD3 policy, learned on simulated or historical risk states, will continue to select appropriate budgets when deployed on live devices facing previously unseen combinations of channel, semantic, and resource conditions.

What would settle it

Observing a significant drop in utility or privacy protection when the system encounters new risk state combinations not seen during offline training would falsify the central claim.

Figures

Figures reproduced from arXiv: 2510.17162 by Guanjie Cheng, Junqin Huang, Linghe Kong, Shiguang Deng, Siyang Liu, Xinkui Zhao, Yishan Chen.

Figure 1
Figure 1. Figure 1: We design an adaptive lightweight privacy-protection agent framework ALPINE. A closed-loop control process: (1) [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: LightAE with block-granularity scaling. descendant variants via knowledge distillation. For each descen￾dant, we conduct performance profiling and construct a tuple set [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: The performance of TD3 model We further examine TD3’s policy behavior under different weight settings of the reward function—privacy gain weight 𝛼 and utility￾loss weight 𝛽, which also serves to validate the feedback module. The privacy budget is inversely proportional to the noise magnitude. As shown in Fig.4, increasing 𝛼 tightens the privacy budget and injects larger noise; increasing 𝛽 relaxes the budg… view at source ↗
Figure 3
Figure 3. Figure 3: The comparison of three RL models. As shown in Fig.3, DDPG exhibits an approximately linear down￾ward trend. SAC contracts too quickly in mid risk regions, yielding a steeper decision boundary. By contrast, TD3 adjusts the policy more smoothly and adaptively across risk levels, responding sensi￾tively to risk changes while maintaining better stability. Regarding loss convergence, DDPG converges steadily; S… view at source ↗
Figure 5
Figure 5. Figure 5: Evaluation in Intel Berkeley Research Lab Sensor [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Evaluation in UK-DALE In UK-DALE: It provides minute-level measurements of whole￾home (aggregate) power and multiple appliance loads with wide dynamic ranges and abrupt power variations for some devices. We assess the privacy–utility trade-off of BLP in a NILM setting and apply BLP only to the aggregate consumption stream. For NILM, we evaluate classification accuracy under varying privacy budgets by predi… view at source ↗
Figure 7
Figure 7. Figure 7: Evaluation in Diabetes 130-US Hospitals dataset [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Large-Scale Deployment of ALPINE processed per unit time. As shown in Fig.8, even under high con￾currency, latency does not deteriorate noticeably, indicating that ALPINE’s computation is well controlled. When the emulated termi￾nal population scales to thousands of nodes, throughput gradually saturates, suggesting that additional optimization of the processing pipeline is necessary under more extreme conc… view at source ↗
read the original abstract

Mobile edge crowdsensing (MECS) enables large-scale real-time sensing services, but its continuous data collection and transmission pipeline exposes terminal devices to dynamic privacy risks. Existing privacy protection schemes in MECS typically rely on static configurations or coarse-grained adaptation, making them difficult to balance privacy, data utility, and device overhead under changing channel conditions, data sensitivity, and resource availability. To address this problem, we propose ALPINE, a lightweight closed-loop framework for adaptive privacy budget allocation in MECS. ALPINE performs multi-dimensional risk perception on terminal devices by jointly modeling channel, semantic, contextual, and resource risks, and maps the resulting risk state to a privacy budget through an offline-trained TD3 policy. The selected budget is then used to drive local differential privacy perturbation before data transmission, while edge-side privacy-utility evaluation provides feedback for policy switching and periodic refinement. In this way, ALPINE forms a terminal-edge collaborative control loop that enables real-time, risk-adaptive privacy protection with low online overhead. Extensive experiments on multiple real-world datasets show that ALPINE achieves a better privacy-utility trade-off than representative baselines, reduces the effectiveness of membership inference, property inference, and reconstruction attacks, and preserves robust downstream task performance under dynamic risk conditions. Prototype deployment further demonstrates that ALPINE introduces only modest runtime overhead on resource-constrained devices.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper proposes ALPINE, a lightweight closed-loop framework for adaptive privacy budget allocation in mobile edge crowdsensing. It performs multi-dimensional risk perception (channel, semantic, contextual, resource) on terminal devices, maps the risk state to a privacy budget via an offline-trained TD3 policy, applies local differential privacy perturbation, and uses edge-side privacy-utility feedback for policy switching and periodic refinement. Experiments on real-world datasets claim improved privacy-utility trade-offs, reduced effectiveness of membership inference, property inference, and reconstruction attacks, robust downstream task performance under dynamic risks, and modest runtime overhead on resource-constrained devices.

Significance. If the empirical claims hold, ALPINE offers a practical advance over static or coarse-grained privacy schemes in MECS by enabling real-time, risk-adaptive protection through terminal-edge collaboration with low online overhead. The integration of TD3 for budget allocation and closed-loop feedback is a clear strength, but the assessment is limited by the absence of quantitative results, error bars, or ablation details in the provided text.

major comments (2)
  1. [Abstract] Abstract: the central claims of better privacy-utility trade-offs and attack resistance rest on high-level assertions without any reported quantitative metrics, confidence intervals, or ablation results, making it impossible to verify the magnitude or statistical significance of the reported improvements.
  2. [Method (TD3 policy and risk state mapping)] TD3 policy training and deployment description: the load-bearing assumption that the offline-trained TD3 policy generalizes to previously unseen joint combinations of channel, semantic, contextual, and resource risk states is not supported by evidence on training distribution coverage or out-of-distribution testing; without this, the closed-loop adaptation claims cannot be substantiated under live conditions.
minor comments (3)
  1. [Risk perception module] Provide explicit definitions and normalization ranges for each component of the multi-dimensional risk state vector.
  2. [Prototype deployment] Include runtime overhead measurements with specific device models and comparison to baselines in the prototype deployment section.
  3. [Experiments] Add a table or figure showing per-dataset numerical results (e.g., utility loss, attack success rates) with error bars for ALPINE versus baselines.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We appreciate the feedback and have carefully considered each point. Below, we provide point-by-point responses and indicate the revisions we plan to make.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claims of better privacy-utility trade-offs and attack resistance rest on high-level assertions without any reported quantitative metrics, confidence intervals, or ablation results, making it impossible to verify the magnitude or statistical significance of the reported improvements.

    Authors: We agree with the referee that the abstract would be strengthened by including specific quantitative metrics to support the claims. In the revised version, we will update the abstract to report key results from our experiments, including specific improvements in privacy-utility trade-offs (e.g., percentage gains over baselines), reductions in attack success rates for membership inference, property inference, and reconstruction attacks, along with any available confidence intervals or statistical significance indicators. This will make the central claims more verifiable. revision: yes

  2. Referee: [Method (TD3 policy and risk state mapping)] TD3 policy training and deployment description: the load-bearing assumption that the offline-trained TD3 policy generalizes to previously unseen joint combinations of channel, semantic, contextual, and resource risk states is not supported by evidence on training distribution coverage or out-of-distribution testing; without this, the closed-loop adaptation claims cannot be substantiated under live conditions.

    Authors: We acknowledge that the current manuscript does not explicitly detail the coverage of the training distribution or provide dedicated out-of-distribution testing results for the TD3 policy. To address this, we will add a new subsection or expand the experimental section to describe the diversity of risk state combinations used in training (drawn from real-world datasets) and include results from out-of-distribution evaluations to demonstrate generalization. This will better substantiate the closed-loop adaptation claims. revision: yes

Circularity Check

0 steps flagged

No circularity: ALPINE applies standard offline TD3 training to risk-to-budget mapping with separate evaluation

full rationale

The derivation chain consists of multi-dimensional risk modeling followed by an offline-trained TD3 policy (a known RL algorithm) that outputs privacy budgets, with edge feedback used only for post-deployment switching and refinement. Training occurs on simulated/historical states and evaluation uses held-out or real-world datasets, so the policy output is not equivalent to its inputs by construction. No self-definitional steps, no fitted parameters renamed as predictions, and no load-bearing self-citations that reduce the central claim to unverified premises. The framework is self-contained against external benchmarks as described in the abstract and reader's assessment.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the assumption that the four risk dimensions can be reliably quantified on-device and that the TD3 policy learned offline transfers to live conditions; no explicit free parameters or invented entities are named in the abstract.

axioms (1)
  • domain assumption Multi-dimensional risk state can be accurately perceived and mapped to a scalar privacy budget by an offline-trained policy.
    Invoked when the paper states that risk perception jointly models channel, semantic, contextual, and resource risks and feeds them to the TD3 policy.

pith-pipeline@v0.9.0 · 5791 in / 1321 out tokens · 21572 ms · 2026-05-18T06:20:37.493640+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages · 3 internal anchors

  1. [1]

    Saeed Ahmed, YoungDoo Lee, Seung-Ho Hyun, and Insoo Koo. 2019. Mitigating the impacts of covert cyber attacks in smart grids via reconstruction of mea- surement data utilizing deep denoising autoencoders.Energies12, 16 (2019), 3091

    work page 2019

  2. [2]

    Albarrak

    Khalied M. Albarrak. 2024. Securing the Future of Web-Enabled IoT: A Critical Analysis of Web of Things Security.Applied Sciences14, 23 (2024). doi:10.3390/ app142310867

    work page 2024

  3. [3]

    Zainab Alwaisi, Tanesh Kumar, Erkki Harjula, and Simone Soderi. 2024. Securing constrained IoT systems: A lightweight machine learning approach for anomaly detection and prevention.Internet of Things28 (2024), 101398. doi:10.1016/j.iot. 2024.101398

    work page doi:10.1016/j.iot 2024

  4. [4]

    Anuj Bhardwaj, Raza Hasan, Shakeel Ahmad, and Salman Mahmood. 2024. Dia- betic Patient Readmission Predictive Analysis: A Comparative Study of Machine Learning Models of Hospital Readmissions. In2024 2nd International Conference on Computing and Data Analytics (ICCDA). IEEE, 1–6

    work page 2024

  5. [5]

    Rui Bi, Meng Zhao, Zhijun Ying, Yiming Tian, and Jianhua Xiong. 2024. Achieving dynamic privacy measurement and protection based on reinforcement learning for mobile edge crowdsensing of IoT.Digital Communications and Networks10, 2 (2024), 380–388

    work page 2024

  6. [6]

    George EP Box, Gwilym M Jenkins, Gregory C Reinsel, and Greta M Ljung. 2015. Time series analysis: forecasting and control. John Wiley & Sons

    work page 2015

  7. [7]

    Mahawaga Arachchige Pathum Chamikara, Peter Bertok, Dongxi Liu, Seyit Camtepe, and Ibrahim Khalil. 2020. Efficient privacy preservation of big data for accurate data mining.Information Sciences527 (2020), 420–443

    work page 2020

  8. [8]

    Rachel Cummings and David Durfee. 2020. Individual sensitivity preprocessing for data privacy. InProceedings of the Fourteenth Annual ACM-SIAM Symposium on Discrete Algorithms. SIAM, 528–547

    work page 2020

  9. [9]

    Yogesh D Deshpande and SR Rahman. 2023. Edge-based real-time sensor data processing for anomaly detection in industrial IoT applications.Research Journal of Computer Systems and Engineering4, 2 (2023), 16–30

    work page 2023

  10. [10]

    Cynthia Dwork, Aaron Roth, et al. 2014. The algorithmic foundations of differ- ential privacy.Foundations and Trends®in Theoretical Computer Science9, 3–4 (2014), 211–407

    work page 2014

  11. [11]

    Z., Song, Z., Woodruff, D

    Shiyuan Feng, Ying Feng, George Z. Li, Zhao Song, David P. Woodruff, and Lichen Zhang. 2025. On Differential Privacy for Adaptively Solving Search Problems via Sketching. arXiv:2506.05503

    work page arXiv 2025

  12. [12]

    Matt Fredrikson, Somesh Jha, and Thomas Ristenpart. 2015. Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures. In Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communica- tions Security(Denver, Colorado, USA)(CCS ’15). Association for Computing Machinery, New York, NY, USA, 1322–1333. doi:10.1145/2810103.2813677

    work page doi:10.1145/2810103.2813677 2015

  13. [13]

    Scott Fujimoto, Herke van Hoof, and David Meger. 2018. Addressing Func- tion Approximation Error in Actor-Critic Methods. InProceedings of the 35th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 80), Jennifer Dy and Andreas Krause (Eds.). PMLR, 1587–1596. https://proceedings.mlr.press/v80/fujimoto18a.html

    work page 2018

  14. [14]

    Gunter, and Nikita Borisov

    Karan Ganju, Qi Wang, Wei Yang, Carl A. Gunter, and Nikita Borisov. 2018. Property Inference Attacks on Fully Connected Neural Networks using Per- mutation Invariant Representations. InProceedings of the 2018 ACM SIGSAC Conference on Computer and Communications Security(Toronto, Canada)(CCS ’18). Association for Computing Machinery, New York, NY, USA, 619...

    work page doi:10.1145/3243734.3243834 2018

  15. [15]

    Ming Gao, Fu Xiao, Weiran Liu, Wentao Guo, Yangtao Huang, Yajie Liu, and Jin- song Han. 2023. Expelliarmus: Command Cancellation Attacks on Smartphones Using Electromagnetic Interference. InProceedings of the IEEE Conference on Computer Communications (INFOCOM). 1–10. doi:10.1109/INFOCOM53939.2023. 10228859

    work page doi:10.1109/infocom53939.2023 2023

  16. [16]

    Quan Geng and Pramod Viswanath. 2016. Optimal Noise Adding Mechanisms for Approximate Differential Privacy.IEEE Transactions on Information Theory 62, 2 (2016), 952–969. doi:10.1109/TIT.2015.2504972

    work page doi:10.1109/tit.2015.2504972 2016

  17. [17]

    Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. 2018. Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor. InProceedings of the 35th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 80), Jennifer Dy and Andreas Krause (Eds.). PMLR, 1861–1870. https://...

    work page 2018

  18. [18]

    Song Han, Jeff Pool, John Tran, and William Dally. 2015. Learning both Weights and Connections for Efficient Neural Network. InAdvances in Neural Information Processing Systems, C. Cortes, N. Lawrence, D. Lee, M. Sugiyama, and R. Garnett (Eds.), Vol. 28. Curran Associates, Inc. https://proceedings.neurips.cc/paper_ files/paper/2015/file/ae0eb3eed39d2bcef4...

    work page 2015

  19. [19]

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the Knowledge in a Neural Network. arXiv:1503.02531 [stat.ML] https://arxiv.org/abs/1503.02531

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  20. [20]

    Jiahui Hu, Jiacheng Du, Zhibo Wang, Xiaoyi Pang, Yajie Zhou, Peng Sun, and Kui Ren. 2024. Does differential privacy really protect federated learning from gradient leakage attacks?IEEE Transactions on Mobile Computing23, 12 (2024), 12635–12649

    work page 2024

  21. [21]

    Weizhe Hua, Yuan Zhou, Christopher M De Sa, Zhiru Zhang, and G Edward Suh

    work page

  22. [22]

    Channel gating neural networks.Advances in neural information processing systems32 (2019)

    work page 2019

  23. [23]

    Kyle Hundman, Valentino Constantinou, Christopher Laporte, Ian Colwell, and Tom Soderstrom. 2018. Detecting spacecraft anomalies using lstms and nonpara- metric dynamic thresholding. InProceedings of the 24th ACM SIGKDD interna- tional conference on knowledge discovery & data mining. 387–395

    work page 2018

  24. [24]

    Md Ruman Islam, Raja Hasnain Anwar, Spyridon Mastorakis, and Muham- mad Taqi Raza. 2024. Characterizing Encrypted Application Traffic Through Cellular Radio Interface Protocol. In2024 IEEE 21st International Conference on Mobile Ad-Hoc and Smart Systems (MASS). IEEE, 321–329

    work page 2024

  25. [25]

    Chenying Jin, Xiang Feng, and Huiqun Yu. 2025. Embracing Multiheterogene- ity and Privacy Security Simultaneously: A Dynamic Privacy-Aware Federated Reinforcement Learning Approach.IEEE Transactions on Neural Networks and Learning Systems36, 5 (2025), 8772–8786. doi:10.1109/TNNLS.2024.3427789

    work page doi:10.1109/tnnls.2024.3427789 2025

  26. [26]

    Bogdan Kulynych, Juan F Gomez, Georgios Kaissis, Flavio du Pin Calmon, and Carmela Troncoso. 2024. Attack-aware noise calibration for differential privacy. Advances in Neural Information Processing Systems37 (2024), 134868–134901

    work page 2024

  27. [27]

    Ninghui Li, Tiancheng Li, and Suresh Venkatasubramanian. 2007. t-Closeness: Privacy Beyond k-Anonymity and l-Diversity. In2007 IEEE 23rd International Conference on Data Engineering. 106–115. doi:10.1109/ICDE.2007.367856

    work page doi:10.1109/icde.2007.367856 2007

  28. [28]

    Pengzhi Li, Yan Pei, and Jianqiang Li. 2023. A comprehensive survey on design and application of autoencoder in deep learning.Applied Soft Computing138 (2023), 110176. doi:10.1016/j.asoc.2023.110176

    work page doi:10.1016/j.asoc.2023.110176 2023

  29. [29]

    Zhuohang Li, Andrew Lowy, Jing Liu, Toshiaki Koike-Akino, Kieran Parsons, Bradley Malin, and Ye Wang. 2024. Analyzing inference privacy risks through gra- dients in machine learning. InProceedings of the 2024 on ACM SIGSAC Conference WWW ’26, April 13–17, 2026, Dubai, United Arab Emirates Trovato et al. on Computer and Communications Security. 3466–3480

    work page 2024

  30. [30]

    Zhetao Li, Xiyu Zeng, Yong Xiao, Chengxin Li, Wentai Wu, and Haolin Liu. 2025. Pattern-Sensitive Local Differential Privacy for Finite-Range Time-Series Data in Mobile Crowdsensing.IEEE Transactions on Mobile Computing24, 1 (2025), 1–14. doi:10.1109/TMC.2024.3445973

    work page doi:10.1109/tmc.2024.3445973 2025

  31. [31]

    Timothy P Lillicrap, Jonathan J Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa, David Silver, and Daan Wierstra. 2015. Continuous control with deep reinforcement learning.arXiv preprint arXiv:1509.02971(2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  32. [32]

    Fei Tony Liu, Kai Ming Ting, and Zhi-Hua Zhou. 2008. Isolation forest. In2008 eighth ieee international conference on data mining. IEEE, 413–422

    work page 2008

  33. [33]

    Vincent B Liu, Laura Y Sue, and Yingnian Wu. 2024. Comparison of machine learning models for predicting 30-day readmission rates for patients with diabetes. Journal of Medical Artificial Intelligence7 (2024)

    work page 2024

  34. [34]

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. 2023. itransformer: Inverted transformers are effective for time series forecasting.arXiv preprint arXiv:2310.06625(2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  35. [35]

    Donghao Luo and Xue Wang. 2024. Moderntcn: A modern pure convolution structure for general time series analysis. InThe twelfth international conference on learning representations. 1–43

    work page 2024

  36. [36]

    Pankaj Malhotra, Lovekesh Vig, Gautam Shroff, Puneet Agarwal, et al. 2015. Long short term memory networks for anomaly detection in time series. InProceedings, Vol. 89. 94

    work page 2015

  37. [37]

    Frank D McSherry. 2009. Privacy integrated queries: an extensible platform for privacy-preserving data analysis. InProceedings of the 2009 ACM SIGMOD International Conference on Management of data. 19–30

    work page 2009

  38. [38]

    Nasr, and Samreen Shahwar

    Khalid Mohiuddin, Huda Fatima, Mohiuddin Ali Khan, Mohammad Abdul Khaleel, Osman A. Nasr, and Samreen Shahwar. 2022. Mobile learning evolution and emerging computing paradigms: An edge-based cloud architecture for reduced latencies and quick response time.Array16 (2022), 100259. doi:10.1016/j.array. 2022.100259

    work page doi:10.1016/j.array 2022

  39. [39]

    Ke Pan and Kaiyuan Feng. 2023. Differential privacy-enabled multi-party learning with dynamic privacy budget allocating strategy.Electronics12, 3 (2023), 658

    work page 2023

  40. [40]

    Danijela D Protić. 2018. Review of KDD Cup ‘99, NSL-KDD and Kyoto 2006+ datasets.Vojnotehnički glasnik/Military Technical Courier66, 3 (2018), 580–596

    work page 2018

  41. [41]

    Protection Regulation. 2018. General data protection regulation.Intouch25 (2018), 1–5

    work page 2018

  42. [42]

    Haoyu Ren, Darko Anicic, and Thomas A. Runkler. 2022. Towards Semantic Management of On-Device Applications in Industrial IoT.ACM Trans. Internet Technol.22, 4, Article 102 (Nov. 2022), 30 pages. doi:10.1145/3510820

    work page doi:10.1145/3510820 2022

  43. [43]

    Wager, Weisong Shi, Timothy Dittrich, Carol Miller, and Kishore Gopalakrishnan

    Javad Roostaei, Yongli Z. Wager, Weisong Shi, Timothy Dittrich, Carol Miller, and Kishore Gopalakrishnan. 2023. IoT-based edge computing (IoTEC) for improved environmental monitoring.Sustainable Computing: Informatics and Systems38 (2023), 100870. doi:10.1016/j.suscom.2023.100870

    work page doi:10.1016/j.suscom.2023.100870 2023

  44. [44]

    2005.Fuzzy logic with engineering applications

    Timothy J Ross. 2005.Fuzzy logic with engineering applications. John Wiley & Sons

    work page 2005

  45. [45]

    Thomas L. Saaty. 1990. How to make a decision: The analytic hierarchy process. European Journal of Operational Research48, 1 (1990), 9–26. doi:10.1016/0377- 2217(90)90057-I Desicion making by the analytic hierarchy process: Theory and applications

    work page doi:10.1016/0377- 1990

  46. [46]

    Thomas L. Saaty. 2004. Fundamentals of the analytic network process — Dependence and feedback in decision-making with a single network.Jour- nal of Systems Science and Systems Engineering13, 2 (April 2004), 129–157. doi:10.1007/s11518-006-0158-y

    work page doi:10.1007/s11518-006-0158-y 2004

  47. [47]

    Michael Schmid, David Rath, and Ulrike Diebold. 2022. Why and how Savitzky– Golay filters should be replaced.ACS Measurement Science Au2, 2 (2022), 185–196

    work page 2022

  48. [48]

    Bernhard Schölkopf, John C Platt, John Shawe-Taylor, Alex J Smola, and Robert C Williamson. 2001. Estimating the support of a high-dimensional distribution. Neural computation13, 7 (2001), 1443–1471

    work page 2001

  49. [49]

    Sheng Shen, Tianqing Zhu, Di Wu, Wei Wang, and Wanlei Zhou. 2022. From distributed machine learning to federated learning: In the view of data privacy and security.Concurrency and Computation: Practice and Experience34, 16 (2022), e6002

    work page 2022

  50. [50]

    Reza Shokri, Marco Stronati, Congzheng Song, and Vitaly Shmatikov. 2017. Mem- bership Inference Attacks Against Machine Learning Models. In2017 IEEE Sym- posium on Security and Privacy (SP). 3–18. doi:10.1109/SP.2017.41

    work page doi:10.1109/sp.2017.41 2017

  51. [51]

    Lisha Shuai, Jiamin Zhang, Yu Cao, Min Zhang, and Xiaolong Yang. 2022. R-DP: A risk-adaptive privacy protection scheme for mobile crowdsensing in industrial internet of things.IET Information Security16, 5 (2022), 373–389

    work page 2022

  52. [52]

    Djordje Slijepčević, Maximilian Henzl, Lukas Daniel Klausner, Tobias Dam, Peter Kieseberg, and Matthias Zeppelzauer. 2021. k-Anonymity in practice: How generalisation and suppression affect machine learning classifiers.Computers & Security111 (2021), 102488. doi:10.1016/j.cose.2021.102488

    work page doi:10.1016/j.cose.2021.102488 2021

  53. [53]

    Ya Su, Youjian Zhao, Chenhao Niu, Rong Liu, Wei Sun, and Dan Pei. 2019. Robust anomaly detection for multivariate time series through stochastic recurrent neural network. InProceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining. 2828–2837

    work page 2019

  54. [54]

    Deepika Suhag and Vivekanand Jha. 2023. A comprehensive survey on mobile crowdsensing systems.Journal of Systems Architecture142 (2023), 102952. doi:10. 1016/j.sysarc.2023.102952

    work page arXiv 2023

  55. [55]

    Zemin Sun, Geng Sun, Long He, Fang Mei, Shuang Liang, and Yanheng Liu

    work page

  56. [56]

    InIEEE INFOCOM 2024-IEEE Conference on Computer Communications

    A two time-scale joint optimization approach for uav-assisted mec. InIEEE INFOCOM 2024-IEEE Conference on Computer Communications. IEEE, 91–100

    work page 2024

  57. [57]

    LATANYA SWEENEY. 2002. k-ANONYMITY: A MODEL FOR PROTECTING PRIVACY.International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems10, 05 (2002), 557–570. arXiv:https://doi.org/10.1142/S0218488502001648 doi:10.1142/S0218488502001648

    work page doi:10.1142/s0218488502001648 2002

  58. [58]

    Richard Yu, Renchao Xie, Azzedine Boukerche, Tao Huang, and Yunjie Liu

    Qinqin Tang, F. Richard Yu, Renchao Xie, Azzedine Boukerche, Tao Huang, and Yunjie Liu. 2022. Internet of Intelligence: A Survey on the Enabling Technologies, Applications, and Challenges.IEEE Communications Surveys & Tutorials24, 3 (2022), 1394–1434. doi:10.1109/COMST.2022.3175453

    work page doi:10.1109/comst.2022.3175453 2022

  59. [59]

    Mengyuan Wang, Hongbo Jiang, Peng Peng, Youhuan Li, and Wenbin Huang

    work page

  60. [60]

    InIEEE INFOCOM 2024 - IEEE Conference on Computer Com- munications

    Toward Accurate Butterfly Counting with Edge Privacy Preserving in Bipartite Networks. InIEEE INFOCOM 2024 - IEEE Conference on Computer Com- munications. 2289–2298. doi:10.1109/INFOCOM52122.2024.10621436

    work page doi:10.1109/infocom52122.2024.10621436 2024

  61. [61]

    Yongfeng Wang, Zheng Yan, Wei Feng, and Shushu Liu. 2020. Privacy protection in mobile crowd sensing: a survey.World Wide Web23, 1 (2020), 421–452

    work page 2020

  62. [62]

    Dingzhu Wen, Yong Zhou, Xiaoyang Li, Yuanming Shi, Kaibin Huang, and Khaled B. Letaief. 2024. A Survey on Integrated Sensing, Communication, and Computation.IEEE Communications Surveys & Tutorials(2024), 1–1. doi:10.1109/COMST.2024.3521498

    work page doi:10.1109/comst.2024.3521498 2024

  63. [63]

    Honghui Xu, Wei Li, Shaoen Wu, Liang Zhao, and Zhipeng Cai. 2024. APOLLO: Differential Private Online Multi-Sensor Data Prediction with Certified Perfor- mance. In2024 IEEE International Conference on Data Mining (ICDM). 530–539. doi:10.1109/ICDM59182.2024.00060

    work page doi:10.1109/icdm59182.2024.00060 2024

  64. [64]

    2010.Distributed Optimization and Games: A Tutorial Overview

    Bo Yang and Mikael Johansson. 2010.Distributed Optimization and Games: A Tutorial Overview. Springer London, London, 109–148. doi:10.1007/978-0-85729- 033-5_4

    work page doi:10.1007/978-0-85729- 2010

  65. [65]

    Yaoqi Yang, Bangning Zhang, Daoxing Guo, Hongyang Du, Zehui Xiong, Dusit Niyato, and Zhu Han. 2024. Generative AI for Secure and Privacy-Preserving Mobile Crowdsensing.IEEE Wireless Communications31, 6 (2024), 29–38. doi:10. 1109/MWC.004.2400017

    work page 2024

  66. [66]

    Obaidat, Shuchen Li, Pengfei Wang, and Kuei-Fang Hsiao

    Ruiyun Yu, Ann Move Oguti, Mohammad S. Obaidat, Shuchen Li, Pengfei Wang, and Kuei-Fang Hsiao. 2023. Blockchain-based solutions for mobile crowdsensing: A comprehensive survey.Computer Science Review50 (2023), 100589. doi:10. 1016/j.cosrev.2023.100589

    work page arXiv 2023

  67. [67]

    Nemer Alberto Zaguir, Guilherme Henrique de Magalhães, and Mauro de Mesquita Spinola. 2024. Challenges and Enablers for GDPR Compliance: System- atic Literature Review and Future Research Directions.IEEE Access12 (2024), 81608–81630. doi:10.1109/ACCESS.2024.3406724

    work page doi:10.1109/access.2024.3406724 2024

  68. [68]

    Qinglong Zhang, Rui Han, Gaofeng Xin, Chi Harold Liu, Guoren Wang, and Lydia Y. Chen. 2022. Lightweight and Accurate DNN-Based Anomaly Detection at Edge.IEEE Transactions on Parallel and Distributed Systems33, 11 (2022), 2927–2942. doi:10.1109/TPDS.2021.3137631

    work page doi:10.1109/tpds.2021.3137631 2022

  69. [69]

    Qiong Zhang, Taochun Wang, Yuan Tao, Nuo Xu, Fulong Chen, and Dong Xie. 2024. Location privacy protection method based on differential pri- vacy in crowdsensing task allocation.Ad Hoc Networks158 (2024), 103464. doi:10.1016/j.adhoc.2024.103464

    work page doi:10.1016/j.adhoc.2024.103464 2024

  70. [70]

    Guangxu Zhu, Zhonghao Lyu, Xiang Jiao, Peixi Liu, Mingzhe Chen, Jie Xu, Shuguang Cui, and Ping Zhang. 2023. Pushing AI to wireless network edge: An overview on integrated sensing, communication, and computation towards 6G. Science China Information Sciences66, 3 (2023), 130301. ALPINE: A Lightweight and Adaptive Privacy-Decision Agent Framework for Dynami...

    work page 2023