Pith. sign in

REVIEW 3 major objections 5 minor 35 references

Profiling Electric Vehicles via Early Charging Voltage Patterns

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Early voltage signals during charging are enough to identify individual electric vehicles, with a random forest reaching 0.86 accuracy on 7,408 sessions from 49 vehicles.

desk verdict Plausible and useful idea, but the 80/20 split is not shown to be session-disjoint, so the headline 0.86 accuracy may be inflated by leakage. read the letter →

arxiv 2506.07714 v1 pith:CAN5AEUF submitted 2025-06-09 cs.CR cs.ETcs.LG

Francesco Marchiori , Denis Donadel , Alessandro Brighente , Mauro Conti This is my paper
classification cs.CRcs.ETcs.LG
keywords electricvehicleprofilingchargingvoltageconstantcurrentphaserandomforestauthenticationprivacystateofchargeside-channelattack
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that an electric vehicle can be identified from voltage measurements taken in the very first phase of charging, long before a full charge completes. Previous profiling methods relied on the charging-current tail at the end of the process, which demands near-complete sessions and gives attackers time to steal energy. The authors test the idea on 7,408 charging sessions from 49 vehicles, extracting statistical and frequency-domain features from early voltage and state-of-charge traces and running five lightweight classifiers. The best result is 0.86 accuracy with a random forest on 128-second windows, and about ten features capture most of that performance. If the claim holds, charging stations could authenticate vehicles within about two minutes and, conversely, anyone with access to voltage traces could track vehicles across locations.

What carries the argument

The object that carries the argument is the early voltage window: 128-second snippets from the start of the constant-current phase, filtered to state of charge at or below 60%. Each window is converted by an automated feature extractor into hundreds of statistical, frequency-domain, and stochastic-dynamics features, and a one-vs-rest random forest then decides if the window belongs to the target vehicle. The named identity in the paper is the partition of the Li-ion charging curve into a constant-current phase, where voltage rises slowly, and a constant-voltage phase, where current falls in a tail; the paper shows that the dynamics of that rise are vehicle-specific.

What would settle it

Locate the data split in the released code and rerun the 0.86-accuracy experiment with a strict session-level split; if accuracy collapses toward chance, the central claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that the constant-current phase of an EV charge, the early stretch usually ignored in favor of the current tail, carries a per-vehicle voltage signature. The authors hypothesize that early voltage behavior is as distinctive as late-stage current behavior, and their experiments support this on a large anonymized dataset. The random forest reached 0.86 accuracy on 128-second windows, with comparable performance from 64 to 256 seconds, and about ten features reproduce most of the result. The authors further show that the classifier still works when state of charge is withheld, and that anonymization by per-segment perturbation does not destroy the signature. In their terms, this establishes a new physical-layer authentication factor and simultaneously broadens the privacy threat from charging data.

Load-bearing premise

The paper's result depends on the train/test split being made by whole charging sessions; if windows from the same session appear on both sides, the model could memorize session-specific voltage drift rather than the vehicle's identity.

Editorial extensions

If this is right

  • Charging columns can begin identity checks about two minutes into a charging session, rather than waiting for the battery to approach full charge.
  • A small set of ten voltage-derived features, mostly frequency-domain and fluctuation statistics, is enough for near-optimal performance, making the classifier light enough for on-station deployment.
  • The method keeps most of its accuracy when state of charge is withheld as a feature, so an adversary with only voltage traces and rough phase information can still profile vehicles.
  • Because the dataset was anonymized by perturbing and interpolating per-segment averages, the results indicate that simple anonymization does not remove the identifying voltage signature.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension the paper does not report is direct multiclass identification: the 0.86 figure comes from one-vs-rest binary tasks, and a 49-way open-set experiment would likely give a lower ceiling for real-world tracking.
  • The paper does not test temporal stability; if fingerprints drift with battery aging, season, or charger model, the authentication and tracking value would shrink, so a longitudinal study would sharpen the claim.
  • If the signature originates in the vehicle's power-electronics dynamics rather than the cell chemistry, the same early-voltage method could transfer to other battery types and charging standards; the current evidence covers only one dataset and three brands.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes an EV-profiling framework based on voltage measurements collected during the early constant-current phase of charging. The authors process the EVBattery dataset by concatenating 128-second snippets into sessions, filtering to SoC ≤60%, extracting tsfresh features from voltage and SoC windows, and training five lightweight classifiers in a one-vs-rest binary setting for each of 49 vehicles. They report a Random Forest accuracy up to 0.86 on 128-second windows, analyze SHAP feature importance, and show that a 10-feature subset gives near-optimal performance. The paper argues that this enables earlier authentication and exposes privacy risks from voltage-only side channels.

Significance. If the reported results hold under a properly session-disjoint evaluation, the paper makes a useful contribution: it demonstrates that early voltage behavior, rather than the final current tail, can support vehicle identification, reducing the time and data required relative to existing current-profile methods. The open-source code, the use of a large public dataset, and the explicit dual-use privacy discussion are strengths. However, the manuscript underspecifies the train/test split unit, which is critical for the validity of every quantitative claim, and the feature-selection protocol risks optimistic bias.

major comments (3)
  1. [4.3 (with 4.1-4.2)] The 80/20 split is not stated to be session-disjoint. Section 4.1 concatenates snippets into sessions and Section 4.2 extracts features per time window, so a single session can produce many windows; a random window-level split will place windows from the same session in both training and test. Because voltage drift and charging-phase dynamics are highly correlated within a session, a Random Forest can memorize session-specific patterns, inflating the accuracy and F1 values in Figs. 4-6 and 8 and undermining the claim that the model profiles a vehicle on a new, unseen charging session. Please specify the split unit, and if it was not session-disjoint, re-run the evaluation with a group-based split by session ID and report the resulting figures.
  2. [5.4] The feature-selection protocol is circular. SHAP importance is computed on 'a subsample of the whole datasets' and the same data appear to be used both to choose the top-k features and to measure the accuracy reported in Fig. 8. This can optimistically bias the 10-feature result. Please move feature selection inside the cross-validation/training folds and report test accuracy for the selected feature count.
  3. [5 (evaluation setup)] The evaluation does not discuss potential confounding by charging station or charger identity. If vehicles in the dataset are not observed across multiple stations, the classifier may learn station-specific voltage signatures rather than vehicle-level fingerprints, which would weaken the cross-location tracking and authentication claims. Please clarify whether station identifiers are available and, if so, add an analysis of cross-station generalization, for example by training on sessions from one set of stations and testing on sessions from unseen stations.
minor comments (5)
  1. [Acknowledgment] The acknowledgment uses 'founded' where 'funded' is intended.
  2. [5.2] Please clarify whether the ratio manipulation is applied before or after the 80/20 split and whether the test set preserves the same ratio; this affects the interpretation of Fig. 5.
  3. [5.4 / Appendix A.2] In Fig. 7, 'FTT Ang.' should be 'FFT Ang.', and 'Max Lang. Pt.' in the figure should match 'Max Langevin point' in Appendix A.2.
  4. [4.1-4.2] Please define explicitly what constitutes a 'time window' in the classification setup (e.g., a 128-second contiguous slice of a session), since the data-length experiments imply variable-length windows from the concatenated sessions.
  5. [2 / 5] A direct quantitative comparison with the EVScout2.0 current-profile baseline would strengthen the claimed advantage over prior work; the current related-work discussion stops short of such a comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical profiling claim is self-contained and does not reduce to its inputs.

full rationale

The paper's central claim is an empirical ML result: voltage windows from the first constant-current phase are used to train lightweight classifiers on a public dataset (EVBattery, 7408 charging sessions from 49 EVs), and accuracy is reported on an 80/20 split. This is not a derivation from first principles that could collapse into its own assumptions; the hypothesis that early voltage resembles later-stage current is tested directly, not imported from a citation. Self-citations ([6], [7], [21]) are used only for inspiration in model selection and as prior work, not as load-bearing justification for the profiling result. No uniqueness theorem or ansatz is smuggled in via citation. The feature-importance analysis (Section 5.4) selects a 10-feature subset via SHAP on a subsample and then evaluates accuracy with that subset; if the same subsample is used for both selection and evaluation, this is an optimistic-bias/leakage concern, but it is not a definitional or self-citational circularity because the accuracy metric is not identical to the SHAP ranking by construction. Similarly, the 80/20 split in Section 4.3 is not explicitly session-disjoint, which raises a possible same-session leakage risk for the validity of the reported accuracy, but this is a methodological validity concern, not a circularity of the kind where a prediction is equivalent to a fitted input. The paper is therefore self-contained against external data and code, and no specific equation or definition makes the central claim reducible to its inputs.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim is empirical and relies on standard ML tools. The main assumptions are domain-level: that the CC-CV model holds, that anonymization keeps discriminative structure, and that the evaluation split avoids temporal leakage. The free parameters are evaluation choices (SoC cutoff, sample threshold, window length, feature count) rather than physical constants, and they directly influence the reported accuracy.

free parameters (4)
  • SoC threshold (60%) = 60%
    Only charging data with state of charge <= 60% are retained, restricting the analysis to the early constant-current phase. The reported accuracy depends on this cutoff (Section 4.1).
  • Minimum sample requirement (100) = 100
    Vehicles with fewer than 100 charging samples are excluded, resulting in the final 49-vehicle dataset. The choice affects the class distribution and the generalizability of the results (Section 4.1).
  • Data length (128 seconds) = 128 s
    The paper reports the best accuracy at a 128-second window length (Section 5.1, Fig. 4). This is selected from a sweep and is not a parameter of the underlying physical model.
  • Number of features (10) = 10
    The near-optimal feature subset is chosen based on SHAP importance on the same data, and the performance of this subset is then reported (Section 5.4, Fig. 8). The selection is data-driven.
assumptions (4)
  • domain assumption Li-ion batteries follow a constant-current/constant-voltage charging profile in which the early constant-current phase contains vehicle-specific voltage characteristics.
    This underlies the entire feature extraction and is introduced in Section 3.1 with Fig. 1.
  • domain assumption The anonymization applied to the EVBattery dataset (perturbation and interpolation) preserves the temporal patterns and correlations needed for classification.
    Explicitly stated in Section 4.1 and Section 5.5; the validity of this assumption is central to the claimed real-world applicability.
  • domain assumption Measured voltage during charging reflects the EV's battery and charging circuit, not primarily the charging column or grid conditions.
    The framework treats voltage as a vehicle fingerprint; if the column dominates the signal, the profiling would not transfer across stations.
  • domain assumption The random 80/20 train/test split is applied at the level of whole charging sessions, preventing windows from the same session from appearing in both partitions.
    Assumed implicitly in Section 4.3; the paper does not state the split unit, and violation of this assumption would inflate accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Profiling Electric Vehicles via Early Charging Voltage Patterns." pith.science (2026). https://pith.science/paper/CAN5AEUF

@misc{pith2026250607714,
  author       = {Pith},
  title        = {Pith review of: Profiling Electric Vehicles via Early Charging Voltage Patterns},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CAN5AEUF}},
  note         = {Machine review of arXiv:2506.07714}
}
read the original abstract

Electric Vehicles (EVs) are rapidly gaining adoption as a sustainable alternative to fuel-powered vehicles, making secure charging infrastructure essential. Despite traditional authentication protocols, recent results showed that attackers may steal energy through tailored relay attacks. One countermeasure is leveraging the EV's fingerprint on the current exchanged during charging. However, existing methods focus on the final charging stage, allowing malicious actors to consume substantial energy before being detected and repudiated. This underscores the need for earlier and more effective authentication methods to prevent unauthorized charging. Meanwhile, profiling raises privacy concerns, as uniquely identifying EVs through charging patterns could enable user tracking. In this paper, we propose a framework for uniquely identifying EVs using physical measurements from the early charging stages. We hypothesize that voltage behavior early in the process exhibits similar characteristics to current behavior in later stages. By extracting features from early voltage measurements, we demonstrate the feasibility of EV profiling. Our approach improves existing methods by enabling faster and more reliable vehicle identification. We test our solution on a dataset of 7408 usable charges from 49 EVs, achieving up to 0.86 accuracy. Feature importance analysis shows that near-optimal performance is possible with just 10 key features, improving efficiency alongside our lightweight models. This research lays the foundation for a novel authentication factor while exposing potential privacy risks from unauthorized access to charging data.

Figures

Figures reproduced from arXiv: 2506.07714 by the authors.

Figure 1
Figure 1. Charging profile of a Li-ion battery [28]. In this paper, we will show that [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of the proposed profiling framework. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Distribution of samples for each car in our dataset after pre-processing. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Average performance of different data lengths (in seconds). [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Average performance of different ratios (representing the number of au [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Average performance of our RF model trained on different subsets of [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Top 10 features by importance extracted by SHAP. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Accuracy of the classifiers trained on different dataset ratios (i.e., num [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

35 extracted references · 33 canonical work pages

  1. [1]

    International Journal of Information Security22(5), 1395–1421 (2023)

    Alcaraz, C., Cumplido, J., Trivino, A.: Ocpp in the spotlight: threats and counter- measures for electric vehicle charging infrastructures 4.0. International Journal of Information Security22(5), 1395–1421 (2023)

    work page 2023

  2. [2]

    IEEE Network34(3), 200–207 (2020) 16 Francesco Marchiori, Denis Donadel, Alessandro Brighente, and Mauro Conti

    Antoun, J., Kabir, M.E., Moussa, B., Atallah, R., Assi, C.: A detailed security assessment of the ev charging ecosystem. IEEE Network34(3), 200–207 (2020) 16 Francesco Marchiori, Denis Donadel, Alessandro Brighente, and Mauro Conti

    work page 2020

  3. [3]

    IEEE Trans- actions on Vehicular Technology70(11), 11338–11351 (2021)

    Babu, P.R., Amin, R., Reddy, A.G., Das, A.K., Susilo, W., Park, Y.: Robust au- thentication protocol for dynamic charging system of electric vehicles. IEEE Trans- actions on Vehicular Technology70(11), 11338–11351 (2021)

    work page 2021

  4. [4]

    In: 28th USENIX Security Symposium (USENIX Security 19)

    Baker,R., Martinovic,I.: Losingthe car keys: WirelessPHY-LayerinsecurityinEV charging. In: 28th USENIX Security Symposium (USENIX Security 19). pp. 407–

    work page

  5. [5]

    Electric Vehicles Security and Privacy: Challenges, Solutions, and Future Needs

    Brighente, A., Conti, M., Donadel, D., Poovendran, R., Turrin, F., Zhou, J.: Elec- tric vehicles security and privacy: Challenges, solutions, and future needs. arXiv preprint arXiv:2301.04587 (2023)

    work page Pith review arXiv 2023

  6. [6]

    0: Electric vehicle profiling through charging profile

    Brighente, A., Conti, M., Donadel, D., Turrin, F.: Evscout2. 0: Electric vehicle profiling through charging profile. ACM Transactions on Cyber-Physical Systems 8(2), 1–24 (2024)

    work page 2024

  7. [7]

    In: Bertino, E., Shulman, H., Waidner, M

    Brighente, A., Conti, M., Sadaf, I.: Tell me how you re-charge, i will tell you where you drove to: Electric vehicles profiling based on charging-current demand. In: Bertino, E., Shulman, H., Waidner, M. (eds.) Computer Security – ESORICS

    work page

  8. [8]

    Neu- rocomputing307, 72–77 (2018).https://doi.org/https://doi.org/10.1016/j

    Christ, M., Braun, N., Neuffer, J., Kempa-Liehr, A.W.: Time Series FeatuRe Ex- traction on basis of Scalable Hypothesis tests (tsfresh – A Python package). Neu- rocomputing307, 72–77 (2018).https://doi.org/https://doi.org/10.1016/j. neucom.2018.03.067

    doi:10.1016/j 2018

Show all 35 references
  1. [9]

    In: European Symposium on Research in Com- puter Security

    Conti, M., Donadel, D., Poovendran, R., Turrin, F.: Evexchange: A relay attack on electric vehicle charging system. In: European Symposium on Research in Com- puter Security. pp. 488–508. Springer (2022)

    2022

  2. [10]

    In: https://www2.deloitte.com/us/en/insights/focus/future-of-mobility/electric- vehicle-trends-2030.html (Jul 2020)

    Deloitte: Electric vehicles: setting a course for 2030. In: https://www2.deloitte.com/us/en/insights/focus/future-of-mobility/electric- vehicle-trends-2030.html (Jul 2020)

    2020

  3. [11]

    In: Inaugural International Symposium on Vehicle Security & Privacy (2023)

    Gangwal, A., Jain, A., Conti, M., et al.: On the feasibility of profiling electric vehicles through charging data. In: Inaugural International Symposium on Vehicle Security & Privacy (2023)

    2023

  4. [12]

    ACM Transactions on Cyber-Physical Systems (TCPS)6(3), 1–26 (2022)

    Ghafouri, M., Kabir, E., Moussa, B., Assi, C.: Coordinated charging and discharg- ing of electric vehicles: A new class of switching attacks. ACM Transactions on Cyber-Physical Systems (TCPS)6(3), 1–26 (2022)

    2022

  5. [13]

    In: 2019 IEEE Green Technologies Conference(GreenTech)

    Gottumukkala, R., Merchant, R., Tauzin, A., Leon, K., Roche, A., Darby, P.: Cyber-physical system security of vehicle charging stations. In: 2019 IEEE Green Technologies Conference(GreenTech). pp. 1–5 (2019)

    2019

  6. [14]

    arXiv preprint arXiv:2201.12358 (2022)

    He, H., Zhang, J., Wang, Y., Jiang, B., Huang, S., Wang, C., Zhang, Y., Xiong, G., Han, X., Guo, D., et al.: Evbattery: A large-scale electric vehicle dataset for battery health and capacity estimation. arXiv preprint arXiv:2201.12358 (2022)

    2022 arXiv

  7. [15]

    IEEE access9, 128069–128094 (2021)

    Khalid, M.R., Khan, I.A., Hameed, S., Asghar, M.S.J., Ro, J.: A comprehensive reviewonstructuraltopologies,powerlevels,energystoragesystems,andstandards for electric vehicle charging stations and their impacts on grid. IEEE access9, 128069–128094 (2021)

    2021

  8. [16]

    In: 2019 IEEE Electrical Power and Energy Conference (EPEC)

    Khan, O.G.M., El-Saadany, E., Youssef, A., Shaaban, M.: Impact of electric ve- hicles botnets on the power grid. In: 2019 IEEE Electrical Power and Energy Conference (EPEC). pp. 1–5. IEEE (2019)

    2019

  9. [17]

    2019 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, Smart- GridComm 2019 pp

    Lee, Z.J., Johansson, D., Low, S.H.: ACN-Sim: An open-source simulator for data- driven electric vehicle charging research. 2019 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, Smart- GridComm 2019 pp. 411–412 (2019).https:...

    2019

  10. [18]

    In: Proceedings of the Tenth ACM International Conference on Future Energy Systems

    Lee, Z.J., Li, T., Low, S.H.: ACN-Data: Analysis and Applications of an Open EV Charging Dataset. In: Proceedings of the Tenth ACM International Conference on Future Energy Systems. p. 139–149. e-Energy ’19, Association for Computing Machinery, New York, NY, USA (2019)

    2019

  11. [19]

    IEEE Transactions on Smart Grid8(5), 2305–2313 (2016)

    Li, H., Dán, G., Nahrstedt, K.: Portunes+: Privacy-preserving fast authentication for dynamic electric vehicle charging. IEEE Transactions on Smart Grid8(5), 2305–2313 (2016)

    2016

  12. [20]

    In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R

    Lundberg, S.M., Lee, S.I.: A unified approach to interpreting model predictions. In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (eds.) Advances in Neural Information Processing Systems 30, pp. 4765–4774. Curran Associates, Inc....

    2017

  13. [21]

    In: Proceedings of the 2023 ACM SIGSAC Confer- ence on Computer and Communications Security

    Marchiori, F., Conti, M.: Your battery is a blast! safeguarding against counterfeit batteries with authentication. In: Proceedings of the 2023 ACM SIGSAC Confer- ence on Computer and Communications Security. pp. 105–119 (2023)

    2023

  14. [22]

    arXiv preprint arXiv:2503.08956 (2025)

    Marchiori, F., Conti, M.: Leaky batteries: A novel set of side-channel attacks on electric vehicles. arXiv preprint arXiv:2503.08956 (2025)

    2025 arXiv

  15. [23]

    Computers and Electrical Engineering119, 109512 (2024)

    Mookherji, S., Odelu, V., Prasath, R.: Secure ultra fast authentication protocol for electric vehicle charging. Computers and Electrical Engineering119, 109512 (2024)

    2024

  16. [24]

    Energies16(3), 1113 (2023)

    Muhammad, Z., Anwar, Z., Saleem, B., Shahid, J.: Emerging cybersecurity and privacy threats to electric vehicles and their impact on human and environmental sustainability. Energies16(3), 1113 (2023)

    2023

  17. [25]

    In: 2018 In- ternational Conference of Electrical and Electronic Technologies for Automotive

    Mültin, M.: Iso 15118 as the enabler of vehicle-to-grid applications. In: 2018 In- ternational Conference of Electrical and Electronic Technologies for Automotive. pp. 1–6 (2018)

    2018

  18. [26]

    In: NDSS (2023)

    Nasr, T., Torabi, S., Bou-Harb, E., Fachkha, C., Assi, C.: Chargeprint: A frame- work for internet-scale discovery and security analysis of ev charging management systems. In: NDSS (2023)

    2023

  19. [27]

    In: 2012 7th IEEE conference on industrial electronics and applications (ICIEA)

    Shen, W., Vo, T.T., Kapoor, A.: Charging algorithms of lithium-ion batteries: An overview. In: 2012 7th IEEE conference on industrial electronics and applications (ICIEA). pp. 1567–1572. IEEE (2012)

    2012

  20. [28]

    ThunderSky: Instruction manual for LFP/LCP/LMP lithium power battery. Tech. rep., Thunder Sky (2007)

    2007

  21. [29]

    secretservice.gov/investigations/skimming(2024)

    United States Secret Service: ATM & POS Skimming.https://www. secretservice.gov/investigations/skimming(2024)

    2024

  22. [30]

    Energy Informatics 5(1), 3 (2022)

    Unterweger, A., Knirsch, F., Engel, D., Musikhina, D., Alyousef, A., de Meer, H.: An analysis of privacy preservation in electric vehicle charging. Energy Informatics 5(1), 3 (2022)

    2022

  23. [31]

    IEEE Journal of Emerging and Selected Topics in Power Elec- tronics9(4), 4639–4657 (2020)

    Ye, J., Guo, L., Yang, B., Li, F., Du, L., Guan, L., Song, W.: Cyber–physical secu- rity of powertrain systems in modern electric vehicles: Vulnerabilities, challenges, and future visions. IEEE Journal of Emerging and Selected Topics in Power Elec- tronics9(4), 4639–4657 (2020)

    2020

  24. [32]

    IEEE Network30(2), 62–66 (2016)

    Zhang, H., Shu, Y., Cheng, P., Chen, J.: Privacy and performance trade-off in cyber-physical systems. IEEE Network30(2), 62–66 (2016)

    2016

  25. [33]

    Zhang, J., Wang, Y., Jiang, B., He, H., Huang, S., Wang, C., Zhang, Y., Han, X., Guo, D., He, G., et al.: Realistic fault detection of li-ion battery via dynamical deep learning. Nature Communications14(1), 5940 (2023) 18 Francesco Marchiori, Denis Donadel, Alessandro Brighent...

    2023

  26. [424]

    org/conference/usenixsecurity19/presentation/baker

    USENIX Association, Santa Clara, CA (Aug 2019),https://www.usenix. org/conference/usenixsecurity19/presentation/baker

    2019

  27. [2021]

    pp. 651–667. Springer International Publishing, Cham (2021)

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.