Pith. sign in

REVIEW 4 major objections 4 minor 41 references

Side-Channel Attacks Survive Noise Cancellation in 3D Printers

T0 review · 4 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A commercial 3D-printer noise-cancellation feature suppresses acoustic side channels but does not eliminate them, and print duration alone identifies the object 64% of the time.

desk verdict First empirical look at a shipped AMNC countermeasure, but the two abstracts tell opposite stories about acoustic leakage; fix that and add a duration control before this is citable. read the letter →

arxiv 2606.13952 v2 pith:J22SQF23 submitted 2026-06-11 cs.CR cs.ETcs.LG

Eric Yocam , Varghese Vaidyan , Micah Flack , Gurcan Comert , Judith L. Mwakalonge This is my paper · ORCID
classification cs.CRcs.ETcs.LG
keywords side-channelattacks3Dprintingactivemotornoisecancellationacousticsidechannelvibrationprintdurationcyber-physicalsecurityadditivemanufacturing
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 asks whether Active Motor Noise Cancellation (AMNC), a noise-reduction feature shipped in commercial fused-deposition-modeling 3D printers, actually closes the acoustic side channel that attackers use to identify printed objects. Its answer is that suppression is real but incomplete: acoustic classification is at chance for short 30-second windows but rises to 27–40% on longer or distributed sampling, and print duration alone identifies the object 64% of the time. Vibration, which AMNC ignores, leaks geometry-correlated information, with a full-sequence temporal model reaching about 61% accuracy. The paper's bounded conclusion is that AMNC raises the observation time an acoustic attacker needs but leaves timing and vibration side channels open, so it is an acoustic-only defense rather than a complete one. This matters because it is the first empirical test of a shipped hardware countermeasure, not a proposed one.

What carries the argument

The argument is carried by a duration-controlled evaluation protocol applied to a public dataset of 144 synchronized acoustic and vibration recordings from two 3D printers of different architectures. Each recording is paired with its own capture, every sample is grouped so a recording never appears in both training and test folds, and significance is checked with Wilson confidence intervals, a label-permutation null, and an order-shuffle control. The order-shuffle control—destroying the segment order of the temporal vibration model and watching accuracy fall from ≈61% to ≈33%—is the mechanism that isolates genuinely sequential, geometry-correlated structure. The second load-bearing mechanism

What would settle it

A replication that holds total print duration fixed across all object classes—by padding G-code or truncating recordings—would settle the claim: if the 27–40% acoustic and ≈61% temporal vibration accuracies collapse to chance, the residual signal is print duration, not geometry.

Watch

Extended reading notes

Core claim

Under the paper's own framing, the central discovery is that a commercial hardware countermeasure against acoustic side channels is only partially effective. AMNC measurably suppresses motor-resonance acoustic emissions, and on short 30-second windows acoustic classification is at chance (11.11%, p=0.188). Yet on a duration-clean 60-second window accuracy reaches 27.08%, and under distributed sampling it reaches 40.28% against an 8.33% baseline—so the channel is suppressed, not eliminated. The strongest discriminator is print duration itself, which classifies the 12 objects at 63.89% accuracy from recording length alone, and the vibration channel carries genuine geometry-correlated temporal

Load-bearing premise

The results stand on the assumption that the residual acoustic signal (27–40%) and the ≈61% temporal vibration accuracy reflect genuine geometry-correlated information rather than the roughly 30-fold variation in print durations across the 12 object classes, which alone classifies at 63.89%.

Editorial extensions

If this is right

  • AMNC raises the observation time an acoustic attacker needs but does not close the acoustic channel; longer windows or distributed sampling still leak object identity.
  • Print duration is an unmitigated timing side channel that AMNC cannot affect; defenders must consider padding or randomizing print schedules to obscure it.
  • Vibration remains a partial, geometry-correlated side channel that AMNC does not address, and a full-sequence temporal model recovers about 61% accuracy from it.
  • The leak is device-specific: classifiers trained on one printer architecture transfer near chance to another, so an attacker must calibrate per device.
  • The demonstrated threat is closed-set identification of known designs, not full geometric reconstruction; reconstruction-grade attacks would likely need the magnetic or power channels AMNC also leaves untouched.

Reading between the lines

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

  • The dominant role of print duration implies a trivial attack: simply measuring how long a print runs identifies the object with high accuracy, and this channel is entirely outside AMNC's scope.
  • The reappearance of acoustic leakage with longer or distributed sampling suggests AMNC's suppression may be effective only within a narrow operating regime; studying how the adaptive filter behaves over time could reveal predictable exposure windows.
  • The manuscript's full-text abstract states AMNC fully neutralizes the acoustic channel while the submitted abstract and duration-controlled results say it does not; the pith here follows the more detailed submitted account, but the discrepancy needs resolution before the headline result is cited.
  • The vibration temporal result suggests any complete defense must address structural vibration—for example, chassis damping or randomized acceleration profiles—since acoustic-only countermeasures leave a measurable side channel.
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

4 major / 4 minor

Summary. The full-text version of this paper claims to provide the first empirical evaluation of Active Motor Noise Cancellation (AMNC) on commercial FDM printers, using a public dataset of 144 synchronized acoustic-vibration recordings from two Bambu Lab printers across 12 object classes. It reports that AMNC fully neutralizes the acoustic channel (12.50%, Wilson 95% CI [8.1, 18.9]), that vibration summary features recover ~31% accuracy, and that a full-sequence temporal model reaches 60.65% accuracy with an order-shuffle control at 32.64%. The submitted abstract, however, reports acoustic classification reaching 27.08% on a duration-clean 60-second window and 40.28% under distributed sampling, and concludes the channel is suppressed, not eliminated. The submitted abstract also reports a duration-only classifier at 63.89%, which is not present in the full text, and a duration-equalized vibration result of 45.83% that is also absent. The paper uses grouped five-fold cross-validation, permutation tests, Wilson intervals, and an order-shuffle control, and releases code. The central contradiction between the two abstracts and the missing duration baseline make the paper's main claims unverifiable as submitted.

Significance. Evaluating a deployed hardware countermeasure on a public dataset is a valuable and timely contribution if the results are correct. The methodology is largely sound: grouping by recording prevents leakage, permutation tests are appropriate for the small sample, and the order-shuffle control is a reasonable check for sequential structure. The manuscript is also commendable for releasing code. However, the contribution cannot be assessed because the acoustic result is reported as both 'fully neutralizes' (full text) and 'suppressed within short windows, not eliminated' (submitted abstract). In addition, the temporal vibration result lacks a duration-only baseline, and the submitted abstract's duration-only accuracy (63.89%) is essentially equal to the full-text temporal model's 60.65%, raising a concrete confound that the body never addresses. These are not presentation issues; they determine whether the paper's headline claims are true.

major comments (4)
  1. [Abstract vs. §6.1] The full-text abstract and §6.1 report acoustic accuracy of 12.50% (Wilson 95% CI [8.1, 18.9]) and state that AMNC 'fully neutralizes the acoustic channel.' The submitted abstract reports 27.08% on a duration-clean 60-second window and 40.28% under distributed sampling against the same 8.33% baseline, and states 'the channel is suppressed within short windows, not eliminated.' These are mutually exclusive conclusions. If the full-text numbers are canonical, the submitted abstract's survival figures are unexplained and the blanket 'fully neutralizes' claim goes beyond the 30-second cap described in §4.3. If the submitted abstract is canonical, the full text's central conclusion is false. The manuscript must be reconciled before the paper's primary claim can be evaluated.
  2. [§6.4, Table 2] The temporal model's 60.65% accuracy is attributed to 'geometry-correlated' sequential structure, but the body never reports a duration-only baseline. Table 2 shows per-object mean durations from 240 s to 7,123 s; the submitted abstract states that recording duration alone classifies at 63.89% and correlates with sliced G-code time at r = 0.907. Because 63.89% and 60.65% are statistically indistinguishable, the entire temporal result may be explained by print duration rather than geometry-related evolution. The order-shuffle control (32.64%) does not remove this confound: shuffled sequences still contain the recording's total length. A duration-only classifier or a duration-equalized temporal analysis (e.g., truncated sequences) is required to support the 'geometry-correlated' claim.
  3. [§6.3] The claim that amplitude features 'stable to recording length (std, RMS) retain 27.08%, ruling out a pure duration artifact' is unsupported by any duration-only comparison. The 27.08% figure is computed on features that are not literally constant across recordings of different lengths, and the submitted abstract's duration-only accuracy of 63.89% makes it possible that a substantial part of this signal is still duration-driven. The body needs either a duration-only baseline or an explicit analysis on equalized observation windows to justify this sentence.
  4. [§6.1] The conclusion that acoustic accuracy is 'indistinguishable from the 8.33% random baseline' is an absence claim, but the paper only shows that the Wilson CI includes 8.33%. With 144 total recordings, the test is underpowered; a CI containing the baseline is not evidence of equivalence. An equivalence test, a report of the minimum detectable effect, or a Bayesian analysis is needed if the paper wishes to claim full neutralization. This is particularly important given the contradictory acoustic numbers in the two abstracts.
minor comments (4)
  1. [Title] The submitted abstract title is 'Side-Channel Attacks Survive Noise Cancellation in 3D Printers' while the full text is 'Side-Channel Attacks Bypass Protection in 3D Printers.' The manuscript should state which is the official title and ensure consistency across versions.
  2. [§9] 'AMNC's does not suppress' should read 'AMNC does not suppress' or 'that AMNC's design does not suppress.'
  3. [§6.3] 'frequency-only features are at chance (6.25%)' reports a value below the 8.33% baseline; consider reporting the permutation p-value to support the 'at chance' characterization.
  4. [Table 2] The table would be easier to interpret with a column for number of recordings per object and total duration; the large mean/min/max span makes the duration confound less obvious than it should be.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the study is empirical classification with grouped cross-validation and permutation tests; no fitted parameter is relabeled as a prediction. The full-text and submitted abstracts contradict each other on the acoustic result, but that is an internal consistency problem, not circularity.

full rationale

Walking the claimed derivation chain, the paper proceeds from a public dataset (Madamopoulos–Tsoutsos, Zenodo DOI 10.5281/zenodo.13329934) through fixed feature extraction (MFCCs, vibration summary statistics, 120-segment temporal sequences), to Random Forest / CNN classifiers, evaluated with grouped five-fold cross-validation, Wilson confidence intervals, label-permutation nulls, and an order-shuffle control. At no point does the paper define an input in terms of the target result, fit a parameter to a subset and then report the same quantity as a prediction, or import a load-bearing conclusion from a self-citation. The authors do not cite their own prior work as evidence; the only external anchor is the independent public dataset. The headline numbers (12.50% acoustic vs 8.33% baseline; 31.25% vibration summary vs permutation p=0.005; 60.65% temporal vs 32.64% shuffled) are empirical outcomes that could have gone differently and are not forced by construction. The paper's own Section 8 limitation—"we report acoustic accuracy under AMNC rather than a causal on/off reduction"—further weakens causal language but does not create circularity. What the manuscript does contain is a serious, non-circular consistency problem: the full-text abstract and Section 6.1 claim AMNC "fully neutralizes" the acoustic channel at 12.50% with a CI containing the 8.33% baseline, while the submitted abstract reports 27.08% on a duration-clean 60-second window and 40.28% under distributed sampling, concluding the channel is "suppressed within short windows, not eliminated." Additionally, the full text's temporal result is labeled "geometry-correlated" without reporting the duration-only baseline (63.89% in the submitted abstract), leaving a plausible duration confound. These contradictions are correctness and framing risks, not reductions of the derivation to its own inputs, so the circularity score remains 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical classification study: the central claims depend on dataset choices and classifier hyperparameters rather than a physical derivation. The most consequential free choice is treating recording duration as an uncontrolled variable; object classes vary by up to two orders of magnitude in print time, and the abstract claims duration alone classifies at 63.89%, a control the body omits. The AMNC-on assumption (no off switch) limits causal interpretation.

free parameters (5)
  • T=120 segments = 120
    Temporal model divides each recording into 120 ordered segments; chosen by hand; accuracy may depend on it.
  • Audio cap 30 s = 30 s
    Acoustic features extracted from the first 30 s of audio; chosen by hand; the abstract suggests longer windows (60 s) change results.
  • Random Forest 200 trees = 200
    Classifier hyperparameter for the summary-feature attack; minor effect on conclusions.
  • Notch filter Q values = 15,30,60
    Simulated acoustic notch bank widths used to test filter sensitivity; chosen by hand.
  • 1D-CNN dilated architecture seeds = 3 seeds
    Temporal model averaged over three seeds; architecture details are not fully specified.
assumptions (4)
  • domain assumption AMNC is active and unchanged throughout all recordings; there is no off switch
    The Limitations section states there is no user-accessible off switch, so the evaluation cannot causally compare AMNC on/off.
  • domain assumption Object print duration (88-11,273 s) is not a class confound; accuracy beyond duration is attributed to geometry
    Table 2 shows classes differ by up to two orders of magnitude in duration; the body's temporal model lacks a duration-only baseline, making this assumption load-bearing.
  • domain assumption iPhone microphone and Teensy 4.0 accelerometer captures are synchronized and representative of a realistic attacker's sensors
    The threat model assumes a passive physically proximate adversary with one or more sensors; the dataset pairing is used as ground truth synchronization.
  • standard math Grouped cross-validation and a 200-sample label-permutation null correctly control for repeated recording structure
    Statistical protocol described in Section 5; standard assumption for this kind of evaluation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Side-Channel Attacks Survive Noise Cancellation in 3D Printers." pith.science (2026). https://pith.science/paper/J22SQF23

@misc{pith2026260613952,
  author       = {Pith},
  title        = {Pith review of: Side-Channel Attacks Survive Noise Cancellation in 3D Printers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J22SQF23}},
  note         = {Machine review of arXiv:2606.13952}
}
read the original abstract

Active Motor Noise Cancellation (AMNC) is a noise-reduction feature shipped in commercial fused deposition modeling (FDM) 3D printers. Because it suppresses the acoustic emissions that side-channel attacks exploit, it has security-relevant side effects, though we find no evidence it was designed as a security control. We present a duration-controlled evaluation of side-channel leakage on AMNC-equipped hardware, using a public dataset of 144 synchronized acoustic and vibration recordings from two Bambu Lab printers across 12 object classes. Spectral analysis confirms suppression is measurably active: the motor-resonance band rises only 4.92 dB above its background-relative baseline during printing. Leakage nonetheless survives it. Acoustic classification is at chance for 30-second windows (11.11%, permutation p = 0.188) but reaches 27.08% on a duration-clean 60-second window and 40.28% under distributed sampling, against an 8.33% baseline: the channel is suppressed within short windows, not eliminated. The dominant discriminator, however, is print duration itself, which classifies at 63.89% (95% CI [55.78, 71.28]) from recording length alone and is validated against sliced G-code print time at r = 0.907. Vibration carries genuine geometry information independent of duration: with the observation window equalized by truncation, classification reaches 45.83% against a 25% baseline (permutation p = 0.023). It nonetheless adds no measurable information beyond duration in paired comparison. The leak is architecture-specific, present on the core-XY device (36.11%) and indistinguishable from chance on the bed-slinger (13.89%), and a consumer handset recovers as much as a mounted accelerometer (25.00% vs 26.39%). Noise cancellation raises the observation time an acoustic attacker requires without eliminating the channel, and leaves print duration entirely untouched.

Figures

Figures reproduced from arXiv: 2606.13952 by the authors.

Figure 1
Figure 1. Accuracy by method. Acoustic sits at chance under AMNC; vibration leaks partially; the full-sequence [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Feature ablation. Magnitude features carry the signal; frequency-only features are at chance. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Top-10 vibration feature importance (Random Forest, mean decrease in impurity). Amplitude statistics [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Full-sequence temporal model. Destroying segment order collapses accuracy, confirming a genuine sequential [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Cross-printer transfer. Vibration leakage is device-specific. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Confusion matrix (pooled vibration). Graded, geometry-plausible confusions across the 12 classes. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 1 canonical work pages

  1. [1]

    Intellectual property protection in additive layer manufacturing: Requirements for secure outsourcing

    Mark Yampolskiy, Jacob Gatlin, Anthony Skjellum, and Yuval Elovici. Intellectual property protection in additive layer manufacturing: Requirements for secure outsourcing. InProceedings of the 4th Program Protection and Reverse Engineering Workshop (PPREW-4). ACM, 2014

    2014

  2. [2]

    King, Jacob Gatlin, Sofia Belikovetsky, Adam Brown, Anthony Skjellum, and Yuval Elovici

    Mark Yampolskiy, Wayne E. King, Jacob Gatlin, Sofia Belikovetsky, Adam Brown, Anthony Skjellum, and Yuval Elovici. Security of additive manufacturing: Attack taxonomy and survey.Additive Manufacturing, 21:431–457, 2018

    2018

  3. [3]

    Acoustic side- channel attacks on additive manufacturing systems

    Mohammad Abdullah Al Faruque, Sujit Rokka Chhetri, Arquimedes Canedo, and Jiang Wan. Acoustic side- channel attacks on additive manufacturing systems. In2016 ACM/IEEE 7th International Conference on Cyber-Physical Systems (ICCPS), pages 1–10. IEEE, 2016

    2016

  4. [4]

    Side channels of cyber-physical systems: Case study in additive manufacturing.IEEE Design & Test, 34(4):18–25, 2017

    Sujit Rokka Chhetri and Mohammad Abdullah Al Faruque. Side channels of cyber-physical systems: Case study in additive manufacturing.IEEE Design & Test, 34(4):18–25, 2017. 9 APREPRINT- JUNE15, 2026

    2017

  5. [5]

    Zeltmann, Nikhil Gupta, Nektarios Georgios Tsoutsos, Michail Maniatakos, Jeyavijayan Rajendran, and Ramesh Karri

    Steven E. Zeltmann, Nikhil Gupta, Nektarios Georgios Tsoutsos, Michail Maniatakos, Jeyavijayan Rajendran, and Ramesh Karri. Manufacturing and security challenges in 3D printing.JOM, 68(7):1872–1881, 2016

    2016

  6. [6]

    Acoustic side- channel attacks on printers

    Michael Backes, Markus Dürmuth, Sebastian Gerling, Manfred Pinkal, and Caroline Sporleder. Acoustic side- channel attacks on printers. InProceedings of the 19th USENIX Security Symposium, pages 307–322, Washington, DC, 2010. USENIX Association

    2010

  7. [7]

    Side-channel analysis of cyber-physical systems: Case study on additive manufacturing

    Sujit Rokka Chhetri, Jiang Wan, and Mohammad Abdullah Al Faruque. Side-channel analysis of cyber-physical systems: Case study on additive manufacturing. In2016 ACM/IEEE 7th International Conference on Cyber- Physical Systems (ICCPS). IEEE, 2016

    2016

  8. [8]

    Cybersecurity of additive manufacturing.IEEE Design & Test, 37(2):77–84, 2020

    Nikhil Gupta, Nektarios Georgios Tsoutsos, and Ramesh Karri. Cybersecurity of additive manufacturing.IEEE Design & Test, 37(2):77–84, 2020

    2020

Show all 41 references
  1. [9]

    Encryption is futile: Reconstructing 3D-printed models using the power side-channel

    Jacob Gatlin, Sofia Belikovetsky, Yuval Elovici, Anthony Skjellum, Joshua Lubell, Paul Witherell, and Mark Yampolskiy. Encryption is futile: Reconstructing 3D-printed models using the power side-channel. InProceedings of the 24th International Symposium on Research in Attacks,...

    2021

  2. [10]

    Amirhossein Jamarani, Yazhou Tu, and Xiali Hei. Practitioner paper: Decoding intellectual property: Acoustic and magnetic side-channel attack on a 3D printer.Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, 622:54–74, 2025

    2025

  3. [11]

    Garza et al

    Marco E. Garza et al. One video to steal them all: 3D-printing IP theft through optical side-channels. In Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security (CCS). ACM, 2025

    2025

  4. [12]

    Active motor noise cancellation – A1 Mini feature overview, 2023

    Bambu Lab. Active motor noise cancellation – A1 Mini feature overview, 2023. Product documentation, https://bambulab.com

    2023

  5. [13]

    Bambu lab adds motor noise cancellation to X1 series 3D printers with new firmware update, 2023.https://www.notebookcheck.net, accessed June 2026

    Notebookcheck. Bambu lab adds motor noise cancellation to X1 series 3D printers with new firmware update, 2023.https://www.notebookcheck.net, accessed June 2026

    2023

  6. [14]

    QuietPrint: Protecting 3D printers against acoustic side-channel attacks

    Seyed Ali Ghazi Asgar and Narasimha Reddy. QuietPrint: Protecting 3D printers against acoustic side-channel attacks. InarXiv preprint arXiv:2602.02198, 2026

    2026 arXiv

  7. [15]

    3D printer audio and vibration side channel dataset for vulnerability research in additive manufacturing security.Data in Brief, 57:111002, 2024

    Christos Madamopoulos and Nektarios Georgios Tsoutsos. 3D printer audio and vibration side channel dataset for vulnerability research in additive manufacturing security.Data in Brief, 57:111002, 2024

    2024

  8. [16]

    3D printer audio and vibration side channels, 2024

    Christos Madamopoulos and Nektarios Georgios Tsoutsos. 3D printer audio and vibration side channels, 2024. Dataset,https://doi.org/10.5281/zenodo.13329934

    2024 doi

  9. [17]

    My smartphone knows what you print: Exploring smartphone-based side-channel attacks against 3D printers

    Sujit Rokka Chhetri, Arquimedes Canedo, and Mohammad Abdullah Al Faruque. My smartphone knows what you print: Exploring smartphone-based side-channel attacks against 3D printers. InProceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security (CCS), pag...

    2016

  10. [18]

    Usefulness of acoustic sounds from 3D printers in an eavesdropping process and reconstruction of printed shapes.Electronics, 9(2):297, 2020

    Ireneusz Kubiak, Artur Przybysz, and Andrzej Sta´nczak. Usefulness of acoustic sounds from 3D printers in an eavesdropping process and reconstruction of printed shapes.Electronics, 9(2):297, 2020

    2020

  11. [19]

    Acoustic emission analysis for additive manufacturing side-channel research.Electronics, 10(3):312, 2021

    Andrzej Sta´nczak, Ireneusz Kubiak, Artur Przybysz, and Anna Witenberg. Acoustic emission analysis for additive manufacturing side-channel research.Electronics, 10(3):312, 2021

    2021

  12. [20]

    Turning hearsay into discovery: Industrial 3D printer side channel information translated to stealing the object design

    Aleksandr Dolgavin, Jacob Gatlin, Moti Yung, and Mark Yampolskiy. Turning hearsay into discovery: Industrial 3D printer side channel information translated to stealing the object design. InarXiv preprint arXiv:2509.18366, 2025

    2025

  13. [21]

    Solewicz, Mark Yampolskiy, Jinghui Toh, and Yuval Elovici

    Sofia Belikovetsky, Yosef A. Solewicz, Mark Yampolskiy, Jinghui Toh, and Yuval Elovici. Digital audio signature for 3D printing integrity.IEEE Transactions on Information Forensics and Security, 14(5):1127–1141, 2018

    2018

  14. [22]

    Introduction to differential power analysis.Journal of Cryptographic Engineering, 1(1):5–27, 2011

    Paul Kocher, Joshua Jaffe, Benjamin Jun, and Pankaj Rohatgi. Introduction to differential power analysis.Journal of Cryptographic Engineering, 1(1):5–27, 2011

    2011

  15. [23]

    RSA key extraction via low-bandwidth acoustic cryptanalysis

    Daniel Genkin, Adi Shamir, and Eran Tromer. RSA key extraction via low-bandwidth acoustic cryptanalysis. In Advances in Cryptology – CRYPTO 2014, pages 444–461. Springer, 2014

    2014

  16. [24]

    SideChannel- 3D: Acoustic, vibration, magnetic, and power side-channel 3D printer dataset

    Nathan Costa, Shih-Yuan Yu, Arnav Malawade, Sujit Chhetri, and Mohammad Abdullah Al Faruque. SideChannel- 3D: Acoustic, vibration, magnetic, and power side-channel 3D printer dataset. InIEEE DataPort, 2021

    2021

  17. [25]

    Electromagnetic side-channel analysis of neural network accelerators.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 42(11):3787–3800, 2023

    Franz Josef Streit, Benedikt Kraß, Lukas Bauer, and Jörg Henkel. Electromagnetic side-channel analysis of neural network accelerators.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 42(11):3787–3800, 2023. 10 APREPRINT- JUNE15, 2026

    2023

  18. [26]

    Forensics of thermal side-channels in additive manufacturing systems

    Mohammad Abdullah Al Faruque, Sujit Rokka Chhetri, Sina Faezi, and Arquimedes Canedo. Forensics of thermal side-channels in additive manufacturing systems. InTechnical Report, University of California Irvine, 2016

    2016

  19. [27]

    Sizhuang Liang and Raheem Beyah. Hiding my real self! protecting intellectual property in additive manufacturing systems against optical side-channel attacks.IEEE Transactions on Information Forensics and Security, 2023

    2023

  20. [28]

    Fix the leak! an information leakage aware secured cyber-physical manufacturing system

    Sujit Rokka Chhetri, Sina Faezi, and Mohammad Abdullah Al Faruque. Fix the leak! an information leakage aware secured cyber-physical manufacturing system. InDesign, Automation & Test in Europe (DATE), pages 1328–1333. IEEE, 2017

    2017

  21. [29]

    Defending side channel attacks in additive manufacturing systems, 2019

    Mohammad Abdullah Al Faruque, Sujit Rokka Chhetri, and Jiang Wan. Defending side channel attacks in additive manufacturing systems, 2019. U.S. Patent US10,212,185 B1

    2019

  22. [30]

    FLAW3D: A trojan-based cyber attack on the physical outcomes of additive manufacturing.IEEE/ASME Transactions on Mechatronics, 27(6):5361–5370, 2022

    Hammond Pearce, Kaushik Yanamandra, Nikhil Gupta, and Ramesh Karri. FLAW3D: A trojan-based cyber attack on the physical outcomes of additive manufacturing.IEEE/ASME Transactions on Mechatronics, 27(6):5361–5370, 2022

    2022

  23. [31]

    A blockchain-based G-code protection approach for cyber-physical security in additive manufacturing.Journal of Computing and Information Science in Engineering, 21(4):041007, 2021

    Zhibo Shi, Chen Kan, Wei Tian, and Chen Liu. A blockchain-based G-code protection approach for cyber-physical security in additive manufacturing.Journal of Computing and Information Science in Engineering, 21(4):041007, 2021

    2021

  24. [32]

    Joshua Brandman, Logan Sturm, Jules White, and Christopher Williams. A physical hash for preventing and detecting cyber-physical attacks in additive manufacturing systems.Journal of Computing and Information Science in Engineering, 23(1):011007, 2023

    2023

  25. [33]

    MitM attacks on intellectual property and integrity of additive manufacturing systems: A security analysis.Computers & Security, 140:103810, 2024

    Hamza Alkofahi, Heba Alawneh, and Anthony Skjellum. MitM attacks on intellectual property and integrity of additive manufacturing systems: A security analysis.Computers & Security, 140:103810, 2024

    2024

  26. [34]

    SOK: 3D printer firmware attacks on fused filament fabrication

    Muhammad Haris Rais, Muhammad Ahsan, and Irfan Ahmed. SOK: 3D printer firmware attacks on fused filament fabrication. InProceedings of the 18th USENIX Conference on Offensive Technologies (WOOT’24). USENIX Association, 2024

    2024

  27. [35]

    Mingjie Wu, Zhuoran Song, and Young B. Moon. Detecting cyber-physical attacks in additive manufacturing using machine learning.Journal of Intelligent Manufacturing, 30(3):1537–1549, 2019

    2019

  28. [36]

    NFTs for 3D models: Sustaining ownership in industry 4.0

    Dimitris Mouris and Nektarios Georgios Tsoutsos. NFTs for 3D models: Sustaining ownership in industry 4.0. IEEE Consumer Electronics Magazine, 13(5):13–22, 2024

    2024

  29. [37]

    Cybersecurity road map for digital manufacturing

    Nektarios Georgios Tsoutsos, Nikhil Gupta, and Ramesh Karri. Cybersecurity road map for digital manufacturing. Computer, 53(9):80–84, 2020

    2020

  30. [38]

    dr0wned – cyber-physical attack with additive manufacturing

    Sofia Belikovetsky, Mark Yampolskiy, Jinghui Toh, Jacob Gatlin, and Yuval Elovici. dr0wned – cyber-physical attack with additive manufacturing. In11th USENIX Workshop on Offensive Technologies (WOOT). USENIX Association, 2017

    2017

  31. [39]

    A survey on machine learning-based side-channel analysis

    Jian Xing, Yanping Zhong, Mi Wen, and Jiankun Hu. A survey on machine learning-based side-channel analysis. IEEE Access, 9:63307–63322, 2021

    2021

  32. [40]

    Steven Davis and Paul Mermelstein. Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences.IEEE Transactions on Acoustics, Speech, and Signal Processing, 28(4):357–366, 1980

    1980

  33. [41]

    Edwin B. Wilson. Probable inference, the law of succession, and statistical inference.Journal of the American Statistical Association, 22(158):209–212, 1927. 11

    1927

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.