Pith. sign in

REVIEW 3 major objections 6 minor 83 references

I Know What You Sync: Covert and Side Channel Attacks on File Systems via syncfs

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read syncfs delay lets any app spy on others' I/O activity

desk verdict Real syncfs leak, well-measured, but the end-to-end attacks hinge on a 120–150 kHz call rate that the paper itself flags as anomalous. read the letter →

arxiv 2411.10883 v2 pith:SAQZMRKX submitted 2024-11-16 cs.CR

Cheng Gu , Yicheng Zhang , Nael Abu-Ghazaleh This is my paper · ORCID
classification cs.CR
keywords syncfssidechannelcovertfilesystemsecuritypagecachetimingcross-containerattackwebsitefingerprintingAndroidapplication
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

This paper claims that the Linux syncfs system call, when invoked by an unprivileged process, flushes not only that process's own dirty data but the file system's entire shared page cache, journal, and inode state. Because the flush work scales with what every other process on the same mount has written, the call's duration becomes a readable timing signal about other domains' I/O. The authors use this signal to build a covert channel reaching 7.61 Kbps on NTFS and three fingerprinting attacks (websites, videos, Android apps) with F1 scores above 90%. The significance is that logical isolation between processes, containers, and across OS boundaries is broken by a standard, unprivileged system call that requires no root permissions.

What carries the argument

The central mechanism is the syncfs system call itself, which synchronizes a whole superblock: it marks dirty inodes, issues write-back for all dirty pages through sync_blockdev_nowait() and sync_blockdev(), and finally issues a device-level flush. Because the page cache, inode cache, and journal are shared across processes and containers, the delay of this single call encodes the write behavior of every other isolation domain sharing the mount. The paper's measurement primitive is the unprivileged high-resolution timer (rdtsc on x86, CNTVCT_EL0 on ARM) wrapped around repeated syncfs calls.

What would settle it

Measure syncfs latency while a victim writes to a file on the same mount, then repeat with the victim writing via O_DIRECT (bypassing the page cache) or to a separate mount; if syncfs delay no longer tracks the victim's writes, the leakage depends on shared-buffer behavior rather than on the flush-all semantics of the call itself.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that syncfs acts as a system-wide flush barrier whose latency is proportional to the volume and kind of dirty state accumulated from all clients of the same file system. Four measured I/O operations (write, write with O_SYNC, ftruncate, rename) each inflate syncfs delay by at least 16 times over baseline, and delay grows linearly with write sizes below the 4 KB page threshold before flattening due to kernel flusher-thread parallelism. Because the victim need not call any synchronization primitive, merely write data, this leakage is richer than the contention-based fsync channels proposed earlier. The paper demonstrates the channel across ext4, ext2, XFS, and NTFS (via WSL), across Docker containers, and on Android, supporting the claim that the flush-all behavior is a general property of shared file system implementations.

Load-bearing premise

The attack requires the attacker to invoke syncfs in a tight loop at very high frequency (about 150,000 times per second on Android and 120 kHz on Linux) without being detected, rate-limited, or throttled, and the victim's writes must go through the shared page cache of a single mounted file system.

Editorial extensions

If this is right

  • An unprivileged attacker on a default Linux or Android install can fingerprint which website, video, or app a victim uses by watching syncfs delay patterns, without any file access or special permissions.
  • The covert channel works across OS isolation boundaries, including Linux on WSL flushing NTFS and across Docker containers that share a disk.
  • Because the victim only performs ordinary writes, defenses that monitor application flush behavior or fsync calls do not detect this leakage.
  • Rate-limiting syncfs degrades rather than eliminates the attack: the paper's own data show Android fingerprinting F1 falling from 93.49% at 150 kHz sampling to 32.99% at 15 kHz and to 8.1% at 15 Hz, near random guessing.
  • Container isolation does not contain the leak, enabling container start/stop detection and a cross-container covert channel with 0.23 Kbps bandwidth at 2.4% error rate.

Reading between the lines

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

  • The flush-all design is a property of shared file system state, so analogous sync primitives on other operating systems (for example, macOS or BSD) may exhibit the same timing leakage even though the paper tests only Linux and NTFS via WSL.
  • An OS redesign that limits syncfs to flushing only the caller's own dirty pages would eliminate the leakage; the paper instead proposes tagging temporary files so their dirty pages are exempt from syncfs flushes.
  • The attack's extreme sampling rate (about 150,000 calls per second on Android) is an observable behavioral signature, suggesting anomaly-based detection that caps syncfs frequency could trade acceptable false positives for a hard ceiling on attack accuracy.
  • The linear write-size inference below 4 KB opens the possibility of finer-grained leakage, such as keystroke timing or fragment-length inference, when a victim writes small chunks at regular intervals.
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 / 6 minor

Summary. The paper identifies the syncfs system call as a timing side-channel vector: because syncfs flushes all dirty buffers of a file system, an unprivileged process can measure syncfs delay to infer the I/O activity of other processes sharing the same file system. The authors demonstrate the underlying leakage with controlled experiments on ext4 and other file systems, build a covert channel over syncfs delay on ext4, ext2, XFS, and NTFS (up to 7.61 Kbps with 1.9% error on NTFS), and present three end-to-end fingerprinting attacks: website fingerprinting and video fingerprinting on Linux, and application fingerprinting on Android, all reporting F1 scores above 90%. They additionally report a cross-container container-detection technique and a cross-container covert channel. The attacks assume a victim and attacker share one mounted file system and that the attacker can invoke syncfs in a tight loop using a high-resolution timer.

Significance. If the results hold, the paper introduces a genuinely new class of OS-level side channels that is distinct from the fsync-contention channels in prior work: the victim only needs to write data, not to call fsync, and the leakage is visible across process and container boundaries. The controlled experiments in Section 3 are carefully designed, and the classifier evaluations use 10-fold cross-validation, which is appropriate for the empirical claims. The cross-file-system covert channel and the container extension broaden the attack surface beyond a single configuration. The main limitation is that the practical claims are contingent on an extremely high syncfs invocation rate, which the paper's own Appendix A.4 flags as anomalous, and on the absence of rate limiting or anomaly detection. Because the central mechanism is plausible and well supported but the operational claims need additional work, the appropriate outcome is major revision.

major comments (3)
  1. [§5.3, §7, Fig. 10, Apps. A.4 and B] The end-to-end side-channel attacks rely on the attacker invoking syncfs at approximately 150,000 Hz on Android (Section 5.3) and 120 kHz on Linux, without being throttled or detected. Figure 10 shows that reducing the Android sampling rate from 150,000 Hz to 15,000 Hz drops application-fingerprinting F1 from 93.49% to 32.99%, and at 15 Hz the F1 is 8.1%, close to random guessing. This is a load-bearing premise: it means the headline >90% F1 results do not hold even under a 10x reduction in call rate. Appendix A.4 explicitly concedes that the 150K calls/sec rate is anomalous and likely detectable, and Appendix B responds only that it is unclear whether a low detection threshold is feasible. The paper should either demonstrate a robust attack at rates that are not trivially distinguishable from normal syncfs usage, or provide and evaluate an explicit rate-adaptation or evasion strategy, or substantially temper the practical-validity claims.
  2. [Table 2, §6.3, Algorithms 1-2] The covert-channel results are reported as single-point estimates of bandwidth and error rate, with no confidence intervals, no number of transmissions, and no variance across runs or reboots. Given the large variance in syncfs latency visible in Tables 1 and 8, and the sensitivity of the bit decision to threshold T and the sender loop counts R and K, the claimed maximum bandwidths (up to 7.61 Kbps) and error rates (down to 0.01%) are not yet shown to be stable. The same applies to the cross-container covert channel in Section 6.3. Please report distributions over repeated transmissions and, if possible, across multiple machines or reboots.
  3. [§5.1, §5.2, §5.3] The classifier evaluations use 10-fold cross-validation on traces collected in a single experimental session (e.g., 100 traces per website, 100 per video, and one trace per Android app launch). Random cross-validation folds can inflate accuracy when consecutive traces share network conditions, browser cache state, or background I/O state. To support the generalization claims, the paper should include a temporal split or a repeated-session evaluation in which training and test traces come from different collection periods.
minor comments (6)
  1. [§5.3, §7] The sampling rate is described as "around 150,000 cycles/sec" in Section 5.3 but as "150,000 Hz" and "150 kHz" in Section 7; please clarify that the quantity is syncfs calls per second, not timer cycles.
  2. [§6, Fig. 9] Figure 9 has no labeled axes or time scale; please add units so the container mount/unmount spikes can be interpreted quantitatively.
  3. [§2.2] The sentence "File system-specific operations depend on the implementation of each file system" is repeated verbatim; please remove the duplicate.
  4. [§3.3] The SNR equation is stated as Var(Signal)/Var(Noise), but the text does not define how Signal and Noise are computed from the syncfs latency traces; please specify the estimation procedure.
  5. [Algorithm 1] The receiver pseudocode does not specify how the ending code is detected or how threshold T is chosen; please make the decoding procedure explicit.
  6. [General] No code or data artifact is referenced; making the collection scripts and de-identified traces available would substantially improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the syncfs leakage is empirically demonstrated and all attack accuracies are measured on held-out traces.

full rationale

The paper's central claim is that syncfs flushes shared dirty buffers and that its latency therefore reflects victim I/O activity. This is established by controlled microbenchmarks (Experiment 1 through Experiment 3, Table 1, Figures 2-3), not derived from the attack results. The end-to-end side-channel attacks are evaluated with 10-fold cross-validation on independently collected traces (Section 5, Tables 3-5 and 7), so the reported F1, precision, and recall are measured performance, not fitted quantities renamed as predictions. The covert channel uses an explicit delay threshold and TSC synchronization, and bandwidth/error rates are measured on actual transmissions (Table 2). Self-citations to prior work by the same group appear only for background material and for the SNR>1 exploitability heuristic (Section 3.3, citing reference [16]); none is load-bearing for the syncfs leakage mechanism, and the STFT/CNN methodology is attributed to external work (references [1], [41]-[43]). Appendix A.4 and Appendix B concede that the roughly 150 kHz syncfs call rate is anomalous and potentially detectable, and Figure 10 quantifies the accuracy collapse at lower rates. This is a robustness and detection-evasion limitation, not a circularity: neither the leakage mechanism nor the classifiers assume the absence of rate limiting. No equation or fitted parameter in the paper is shown to be equivalent to the target result by construction, and no central premise is justified only by a self-citation chain. Therefore no significant circularity is present.

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

The paper's claims rest on standard kernel semantics (shared page cache, syncfs flushing the superblock's dirty pages) and on the attacker's unprivileged access to syncfs and timers. No new physical or software entities are introduced. Several empirical parameters (thresholds, sampling rates, ML hyperparameters) are tuned to the experimental setup.

free parameters (5)
  • Covert channel delay threshold T = not disclosed
    Algorithm 1 uses threshold T to classify syncfs delay as bit 1 or 0; T is implicitly calibrated to the observed baseline and signal delays, but its value and selection rule are not given.
  • Sender NOP loop counts R and K = not disclosed
    Algorithm 2 controls bit timing via R (write) and K (no-op) loop iterations; these must be tuned to produce distinguishable syncfs delay patterns, but the values are not reported.
  • Profiler sampling rate = 150,000 Hz (Android), 120 kHz (Linux)
    The attacker calls syncfs at a fixed high rate; the rate is chosen as a trade-off between time resolution and detectability, and the paper shows accuracy drops sharply at lower rates (Figure 10).
  • STFT window size = 256
    STFT window size is a hyperparameter chosen for frequency feature extraction in all three fingerprinting attacks.
  • CNN training hyperparameters = learning rate 1e-4, weight decay 1e-5, batch size 80 (web/app) or 20 (video), 10-fold cross-validation
    ResNet-152 fine-tuning settings are chosen manually and are not derived from the problem.
assumptions (4)
  • domain assumption syncfs flushes all dirty pages and metadata associated with the superblock, including pages dirtied by other processes on the same file system.
    This kernel behavior is the foundation of the leakage. Stated in Section 2.2 and demonstrated indirectly in Section 3 experiments.
  • domain assumption The page cache and journal/inode caches are shared across isolation domains (processes, containers) using the same file system.
    Threat model in Section 2.3 and used in the cross-container attack (Section 6); without shared caches, syncfs delay would not reflect victim activity.
  • domain assumption An unprivileged process can call syncfs and can access a high-resolution timestamp counter without special permissions.
    Required by the threat model and all experiments; true on Linux 2.6.39+, Android 9+, and WSL, as stated in Section 2.3.
  • domain assumption Each target activity (website, video, app) produces a learnable and repeatable syncfs delay pattern in a given environment.
    The fingerprinting attacks assume the STFT/CNN classifiers can separate classes based on consistent write patterns; the paper validates this empirically but does not establish environmental generalization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of I Know What You Sync: Covert and Side Channel Attacks on File Systems via syncfs." pith.science (2026). https://pith.science/paper/SAQZMRKX

@misc{pith2026241110883,
  author       = {Pith},
  title        = {Pith review of: I Know What You Sync: Covert and Side Channel Attacks on File Systems via syncfs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SAQZMRKX}},
  note         = {Machine review of arXiv:2411.10883}
}
read the original abstract

Operating Systems enforce logical isolation using abstractions such as processes, containers, and isolation technologies to protect a system from malicious or buggy code. In this paper, we show new types of side channels through the file system that break this logical isolation. The file system plays a critical role in the operating system, managing all I/O activities between the application layer and the physical storage device. We observe that the file system implementation is shared, leading to timing leakage when using common I/O system calls. Specifically, we found that modern operating systems take advantage of any flush operation (which saves cached blocks in memory to the SSD or disk) to flush all of the I/O buffers, even those used by other isolation domains. Thus, by measuring the delay of syncfs, the attacker can infer the I/O behavior of victim programs. We then demonstrate a syncfs covert channel attack on multiple file systems, including both Linux native file systems and the Windows file system, achieving a maximum bandwidth of 5 Kbps with an error rate of 0.15% on Linux and 7.6 Kbps with an error rate of 1.9% on Windows. In addition, we construct three side-channel attacks targeting both Linux and Android devices. On Linux devices, we implement a website fingerprinting attack and a video fingerprinting attack by tracking the write patterns of temporary buffering files. On Android devices, we design an application fingerprinting attack that leaks application write patterns during boot-up. The attacks achieve over 90% F1 score, precision, and recall. Finally, we demonstrate that these attacks can be exploited across containers implementing a container detection technique and a cross-container covert channel attack.

Figures

Figures reproduced from arXiv: 2411.10883 by the authors.

Figure 1
Figure 1. Overview of Linux file system and I/O system [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. syncfs delay of concurrent I/O system calls on ext4 as this call primarily affects the journal cache. Similarly, the latency of syncfs shows a modest increase for concurrent ftruncate operations, which generate additional dirty journal and inode entries for flushing. In contrast, the write opera￾tion, which affects the journal, inode, and page cache, causes a steep linear increase in syncfs latency as the number of … view at source ↗
Figure 3
Figure 3. syncfs latency with different write size: (a) Below page size (4KB), (b) Above page size (4KB). As the write size increases within the page size limit (4 KB), the syncfs latency grows linearly. This is because larger write sizes generate more dirty page cache entries, journal updates, and inode changes, leading to longer flush￾ing times for syncfs. Specifically, syncfs performs I/O in two stages: sync_blockdev_nowai… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The syncfs covert channel uses TSC for synchro￾nization on ext4 before transmission. Covert channel synchronization. A key component is the synchronization between the sender and receiver, as it significantly enhances bandwidth and reduces the error rate. We follow the…
Figure 5
Figure 5. Figure 5: The syncfs covert channel transmission has a clear difference between bit ’0’ and ’1’ on ext4 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The STFTs of side-channel traces collected during the victim’s access to different websites show distinct patterns [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: The STFTs of side-channel traces collected during the victim’s playback of different videos on [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: The STFTs of side-channel traces collected during the victim’s launch of different applications on an Android [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Detecting container startup and stop using [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Application fingerprint attack results with limited [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

83 extracted references · 74 canonical work pages

  1. [1]

    Idleleak: Exploit- ing idle state side effects for information leakage,

    F. Rauscher, A. Kogler, J. Juffinger, and D. Gruss, “Idleleak: Exploit- ing idle state side effects for information leakage,” in Network and Distributed System Security (NDSS) Symposium 2024 , 2024

    2024

  2. [2]

    Papp: Prefetcher-aware prime and probe side-channel attack,

    D. Wang, Z. Qian, N. Abu-Ghazaleh, and S. V . Krishnamurthy, “Papp: Prefetcher-aware prime and probe side-channel attack,” in Proceedings of the 56th Annual Design Automation Conference 2019, 2019, pp. 1–6

    2019

  3. [3]

    Last-level cache side-channel attacks are practical,

    F. Liu, Y . Yarom, Q. Ge, G. Heiser, and R. B. Lee, “Last-level cache side-channel attacks are practical,” in 2015 IEEE symposium on security and privacy . IEEE, 2015, pp. 605–622

    2015

  4. [4]

    Going through the motions:AR/VR keylogging from user head motions,

    C. Slocum, Y . Zhang, N. Abu-Ghazaleh, and J. Chen, “Going through the motions:AR/VR keylogging from user head motions,” in 32nd USENIX Security Symposium (USENIX Security 23) , 2023, pp. 159– 174

    work page 2023

  5. [5]

    Plac- eraider: Virtual theft in physical spaces with smartphones,

    R. Templeman, Z. Rahman, D. Crandall, and A. Kapadia, “Plac- eraider: Virtual theft in physical spaces with smartphones,” arXiv preprint arXiv:1209.5982, 2012

    arXiv 2012

  6. [6]

    Iner- tiear: Automatic and device-independent IMU-based eavesdropping on smartphones,

    M. Gao, Y . Liu, Y . Chen, Y . Li, Z. Ba, X. Xu, and J. Han, “Iner- tiear: Automatic and device-independent IMU-based eavesdropping on smartphones,” in IEEE INFOCOM 2022-IEEE Conference on Computer Communications. IEEE, 2022, pp. 1129–1138

    work page 2022

  7. [7]

    DRAMA: Exploiting DRAM addressing for cross-CPU attacks,

    P. Pessl, D. Gruss, C. Maurice, M. Schwarz, and S. Mangard, “DRAMA: Exploiting DRAM addressing for cross-CPU attacks,” in 25th USENIX security symposium (USENIX security 16) , 2016, pp. 565–581

    work page 2016

  8. [8]

    DramaQueen: Revisiting side channels in DRAM,

    V . van der Veen and B. Gras, “DramaQueen: Revisiting side channels in DRAM,” 2023

    work page 2023

Show all 83 references
  1. [9]

    Invisible probe: Timing attacks with PCIe congestion side-channel,

    M. Tan, J. Wan, Z. Zhou, and Z. Li, “Invisible probe: Timing attacks with PCIe congestion side-channel,” in 2021 IEEE Symposium on Security and Privacy (SP) . IEEE, 2021, pp. 322–338

    2021

  2. [10]

    Beyond the bridge: Contention-based covert and side channel attacks on multi-GPU interconnect,

    Y . Zhang, R. Nazaraliyev, S. B. Dutta, N. Abu-Ghazaleh, A. Marquez, and K. Barker, “Beyond the bridge: Contention-based covert and side channel attacks on multi-GPU interconnect,” arXiv preprint arXiv:2404.03877, 2024

    2024 arXiv

  3. [11]

    Rendered insecure: GPU side channel attacks are practical,

    H. Naghibijouybari, A. Neupane, Z. Qian, and N. Abu-Ghazaleh, “Rendered insecure: GPU side channel attacks are practical,” in Proceedings of the 2018 ACM SIGSAC conference on computer and communications security, 2018, pp. 2139–2153

    2018

  4. [12]

    Leaky DNN: Stealing deep-learning model secret with GPU context-switching side-channel,

    J. Wei, Y . Zhang, Z. Zhou, Z. Li, and M. A. Al Faruque, “Leaky DNN: Stealing deep-learning model secret with GPU context-switching side-channel,” in 2020 50th Annual IEEE/IFIP International Con- ference on Dependable Systems and Networks (DSN) . IEEE, 2020, pp. 125–137

    2020

  5. [13]

    OS-level side channels without procfs: Exploring cross-App information leakage on iOS,

    X. Zhang, X. Wang, X. Bai, Y . Zhang, and X. Wang, “OS-level side channels without procfs: Exploring cross-App information leakage on iOS,” in Proceedings of the Symposium on Network and Distributed System Security, 2018

    2018

  6. [14]

    The danger of minimum exposures: Understanding cross-App information leaks on iOS through multi-side-channel learning,

    Z. Wang, J. Guan, X. Wang, W. Wang, L. Xing, and F. Alharbi, “The danger of minimum exposures: Understanding cross-App information leaks on iOS through multi-side-channel learning,” in Proceedings of the 2023 ACM SIGSAC Conference on Computer and Communica- tions Security, 202...

    2023

  7. [15]

    No pardon for the interruption: New inference attacks on android through interrupt timing analysis,

    W. Diao, X. Liu, Z. Li, and K. Zhang, “No pardon for the interruption: New inference attacks on android through interrupt timing analysis,” in 2016 IEEE Symposium on Security and Privacy (SP). IEEE, 2016, pp. 414–432

    2016

  8. [16]

    It’s all in your head (set): Side-channel attacks on AR/VR systems,

    Y . Zhang, C. Slocum, J. Chen, and N. Abu-Ghazaleh, “It’s all in your head (set): Side-channel attacks on AR/VR systems,” in32nd USENIX Security Symposium (USENIX Security 23) , 2023, pp. 3979–3996

    2023

  9. [17]

    Write+Sync: Software cache write covert channels exploiting memory-disk synchronization,

    C. Chen, J. Cui, G. Qu, and J. Zhang, “Write+Sync: Software cache write covert channels exploiting memory-disk synchronization,” Trans. Info. For. Sec. , vol. 19, p. 8066–8078, Jan. 2024. [Online]. Available: https://doi.org/10.1109/TIFS.2024.3414255

    2024

  10. [18]

    Sync+Sync: A covert channel built on fsync with storage,

    Q. Jiang and C. Wang, “Sync+Sync: A covert channel built on fsync with storage,” in 33rd USENIX Security Symposium (USENIX Security 24) . Philadelphia, PA: USENIX Association, Aug. 2024, pp. 3349–3366. [Online]. Available: https://www.usenix. org/conference/usenixsecurity24/pr...

    2024

  11. [19]

    R. H. Arpaci-Dusseau and A. C. Arpaci-Dusseau, Operating Systems: Three Easy Pieces . North Charleston, SC, USA: CreateSpace Independent Publishing Platform, 2018

    2018

  12. [20]

    A better update policy,

    J. C. Mogul, “A better update policy,” in USENIX Summer, vol. 94, 1994

    1994

  13. [21]

    Disk scheduling revisited,

    M. Seltzer, P. Chen, and J. Ousterhout, “Disk scheduling revisited,” in Proceedings of the winter 1990 USENIX technical conference . Washington, DC, 1990, pp. 313–323

    1990

  14. [22]

    UBC: An efficient unified I/O and memory caching sub- system for NetBSD,

    C. Silvers, “UBC: An efficient unified I/O and memory caching sub- system for NetBSD,” in 2000 USENIX Annual Technical Conference (USENIX ATC 00) , 2000

    2000

  15. [23]

    Optimistic crash consistency,

    V . Chidambaram, T. S. Pillai, A. C. Arpaci-Dusseau, and R. H. Arpaci-Dusseau, “Optimistic crash consistency,” in Proceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles, 2013, pp. 228–243

    2013

  16. [24]

    Consistency without ordering

    V . Chidambaram, T. Sharma, A. C. Arpaci-Dusseau, and R. H. Arpaci-Dusseau, “Consistency without ordering.” in FAST, vol. 12, 2012, pp. 101–116

    2012

  17. [25]

    Metadata update performance in file systems

    G. R. Ganger and Y . N. Patt, “Metadata update performance in file systems.” in OSDI, vol. 94, 1994, p. 5

    1994

  18. [26]

    Reimplementing the cedar file system using logging and group commit,

    R. Hagmann, “Reimplementing the cedar file system using logging and group commit,” in Proceedings of the eleventh ACM Symposium on Operating systems principles , 1987, pp. 155–162

    1987

  19. [27]

    Houdini’s escape: Breaking the resource rein of Linux control groups,

    X. Gao, Z. Gu, Z. Li, H. Jamjoom, and C. Wang, “Houdini’s escape: Breaking the resource rein of Linux control groups,” in Proceedings of the 2019 ACM SIGSAC Conference on Computer and Communi- cations Security, 2019, pp. 1073–1086

    2019

  20. [28]

    Disk storage isolation and verification in cloud,

    Z. Wang, K. Sun, S. Jajodia, and J. Jing, “Disk storage isolation and verification in cloud,” in 2012 IEEE Global Communications Conference (GLOBECOM). IEEE, 2012, pp. 771–776

    2012

  21. [29]

    SCFS: A shared cloud-backed file system

    A. N. Bessani, R. Mendes, T. Oliveira, N. F. Neves, M. Correia, M. Pasin, and P. Verissimo, “SCFS: A shared cloud-backed file system.” in USENIX Annual Technical Conference . Philadelphia, USA, 2014, pp. 169–180

    2014

  22. [30]

    Ext4: The next generation of ext2/3 filesystem

    M. Cao, S. Bhattacharya, and T. Ts’o, “Ext4: The next generation of ext2/3 filesystem.” in LSF, 2007

    2007

  23. [31]

    Scalability in the xfs file system

    A. Sweeney, D. Doucette, W. Hu, C. Anderson, M. Nishimoto, and G. Peck, “Scalability in the xfs file system.” in USENIX Annual Technical Conference, vol. 15, 1996

    1996

  24. [32]

    BTRFS: The Linux B-tree filesystem,

    O. Rodeh, J. Bacik, and C. Mason, “BTRFS: The Linux B-tree filesystem,” ACM Transactions on Storage (TOS) , vol. 9, no. 3, pp. 1–32, 2013

    2013

  25. [33]

    File system design for an nfs file server appliance

    D. Hitz, J. Lau, and M. A. Malcolm, “File system design for an nfs file server appliance.” in USENIX winter, vol. 94, 1994, pp. 10–5555

    1994

  26. [34]

    CNTVCT EL0, Counter-timer Vir- tual Count register,

    ARM, “CNTVCT EL0, Counter-timer Vir- tual Count register,” https://developer.arm.com/ documentation/ddi0601/2021-12/AArch64-Registers/ CNTVCT-EL0--Counter-timer-Virtual-Count-register, 2024

    2021

  27. [35]

    Documentation for /proc/sys/vm/,

    L. Torvalds, “Documentation for /proc/sys/vm/,” https://github.com/ torvalds/linux/blob/master/Documentation/admin-guide/sysctl/vm.rst, 2024

    2024

  28. [36]

    Low latency performance tuning for red hat enterprise Linux 7,

    J. Mario and J. Eder, “Low latency performance tuning for red hat enterprise Linux 7,” 2017

    2017

  29. [37]

    Hardware countermeasures against DPA–a statistical analysis of their effectiveness,

    S. Mangard, “Hardware countermeasures against DPA–a statistical analysis of their effectiveness,” in Topics in Cryptology–CT-RSA 2004: The Cryptographers’ Track at the RSA Conference 2004, San Francisco, CA, USA, February 23-27, 2004, Proceedings . Springer, 2004, pp. 222–235

  30. [38]

    Levenshtein dis- tance: Information theory, computer science, string (computer sci- ence), string metric, damerau? levenshtein distance, spell checker, hamming distance,

    F. P. Miller, A. F. Vandome, and J. McBrewster, “Levenshtein dis- tance: Information theory, computer science, string (computer sci- ence), string metric, damerau? levenshtein distance, spell checker, hamming distance,” 2009

    2009

  31. [39]

    Barnes, Pro Windows Subsystem for Linux (WSL)

    H. Barnes, Pro Windows Subsystem for Linux (WSL). Springer, 2021

    2021

  32. [40]

    Top 1M sites (2024),

    Alexa, “Top 1M sites (2024),” https://www.alexa.com/topsites

    2024

  33. [41]

    Signal estimation from modified short-time fourier transform,

    D. Griffin and J. Lim, “Signal estimation from modified short-time fourier transform,” IEEE Transactions on acoustics, speech, and signal processing, vol. 32, no. 2, pp. 236–243, 1984

    1984

  34. [42]

    ECG arrhythmia clas- sification using STFT-based spectrogram and convolutional neural network,

    J. Huang, B. Chen, B. Yao, and W. He, “ECG arrhythmia clas- sification using STFT-based spectrogram and convolutional neural network,” IEEE access, vol. 7, pp. 92 871–92 880, 2019

    2019

  35. [43]

    Fingerprint enhancement using STFT analysis,

    S. Chikkerur, A. N. Cartwright, and V . Govindaraju, “Fingerprint enhancement using STFT analysis,” Pattern recognition , vol. 40, no. 1, pp. 198–211, 2007

    2007

  36. [44]

    SciPy 1.0: fundamental algorithms for scientific computing in Python,

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright et al., “SciPy 1.0: fundamental algorithms for scientific computing in Python,” Nature methods, vol. 17, no. 3, pp. 261–272, 2020

    2020

  37. [45]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  38. [46]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga et al. , “Pytorch: An imperative style, high-performance deep learning library,” Advances in neural information processing systems , vol. 32, 2019

    2019

  39. [47]

    A study of cross-validation and bootstrap for accuracy estimation and model selection,

    R. Kohavi, “A study of cross-validation and bootstrap for accuracy estimation and model selection,” Morgan Kaufman Publishing, 1995

    1995

  40. [48]

    WebGPU-SPY: Finding fingerprints in the sandbox through GPU cache attacks,

    E. Ferguson, A. Wilson, and H. Naghibijouybari, “WebGPU-SPY: Finding fingerprints in the sandbox through GPU cache attacks,” arXiv preprint arXiv:2401.04349 , 2024

    2024 arXiv

  41. [49]

    Rendering contention channel made practical in web browsers,

    S. Wu, J. Yu, M. Yang, and Y . Cao, “Rendering contention channel made practical in web browsers,” in31st USENIX Security Symposium (USENIX Security 22) , 2022, pp. 3183–3199

    2022

  42. [50]

    Video content authentication tech- niques: a comprehensive survey,

    R. D. Singh and N. Aggarwal, “Video content authentication tech- niques: a comprehensive survey,” Multimedia Systems , vol. 24, pp. 211–240, 2018

    2018

  43. [51]

    Walls have ears: Traffic-based side-channel attack in video streaming,

    J. Gu, J. Wang, Z. Yu, and K. Shen, “Walls have ears: Traffic-based side-channel attack in video streaming,” in IEEE INFOCOM 2018- IEEE conference on computer communications . IEEE, 2018, pp. 1538–1546

    2018

  44. [52]

    Camcorder recording robust video fingerprinting,

    A. Garboan, M. Mitrea, and F. Preteux, “Camcorder recording robust video fingerprinting,” in 2012 IEEE 16th International Symposium on Consumer Electronics. IEEE, 2012, pp. 1–4

    2012

  45. [53]

    Droppix: Towards more realistic video fingerprinting,

    P. Bła ´skiewicz, M. Klonowski, and P. Syga, “Droppix: Towards more realistic video fingerprinting,” in International Conference on E-Business and Telecommunication Networks , 2020. [Online]. Available: https://api.semanticscholar.org/CorpusID:220847103

    2020

  46. [54]

    Converting video formats with FFmpeg,

    S. Tomar, “Converting video formats with FFmpeg,” Linux journal , vol. 2006, no. 146, p. 10, 2006

    2006

  47. [55]

    Multi-platform distribution of video content: An analysis of video content cross-posted by YouTubers on Bilibili,

    H. Zhang and K. Scheibe, “Multi-platform distribution of video content: An analysis of video content cross-posted by YouTubers on Bilibili,” in International Conference on Human-Computer Interac- tion. Springer, 2023, pp. 149–156

    2023

  48. [56]

    Amazon elastic container service (Amazon ECS),

    Amazon, “Amazon elastic container service (Amazon ECS),” https: //aws.amazon.com/pm/ecs/?nc1=h ls, 2024

    2024

  49. [57]

    Containers on compute engine,

    Google, “Containers on compute engine,” https://cloud.google.com/ compute/docs/containers, 2024

    2024

  50. [58]

    Can applications recover from fsync failures?

    A. Rebello, Y . Patel, R. Alagappan, A. C. Arpaci-Dusseau, and R. H. Arpaci-Dusseau, “Can applications recover from fsync failures?” ACM Transactions on Storage (TOS) , vol. 17, no. 2, pp. 1–30, 2021

    2021

  51. [59]

    Failure-Atomic updates of application data in a Linux file system,

    R. Verma, A. A. Mendez, S. Park, S. S. Mannarswamy, T. P. Kelly, and C. B. Morrey III, “Failure-Atomic updates of application data in a Linux file system,” in 13th USENIX Conference on File and Storage Technologies (FAST 15), 2015, pp. 203–211

    2015

  52. [60]

    PUF- based encryption method for IC cards on-chip memories,

    Y . Bai, L. Wu, X. Wu, X. Li, X. Zhang, and B. Wang, “PUF- based encryption method for IC cards on-chip memories,” Electronics Letters, vol. 52, no. 20, pp. 1671–1673, 2016

    2016

  53. [61]

    A novel key generation scheme using quaternary puf responses and wiretap polar coding,

    Y . Bai and Z. Yan, “A novel key generation scheme using quaternary puf responses and wiretap polar coding,” IEEE Communications Letters, vol. 25, no. 7, pp. 2142–2145, 2021

    2021

  54. [62]

    SoK: Understanding design choices and pitfalls of trusted execution environments,

    M. Li, Y . Yang, G. Chen, M. Yan, and Y . Zhang, “SoK: Understanding design choices and pitfalls of trusted execution environments,” in Proceedings of the 19th ACM Asia Conference on Computer and Communications Security, 2024, pp. 1600–1616

    2024

  55. [63]

    Keydrown: Eliminating software- based keystroke timing side-channel attacks,

    M. Schwarz, M. Lipp, D. Gruss, S. Weiser, C. L. N. Maurice, R. Spreitzer, and S. Mangard, “Keydrown: Eliminating software- based keystroke timing side-channel attacks,” in Network and Dis- tributed System Security Symposium 2018 , 2018, p. 15

    2018

  56. [64]

    Timing side-channel attacks and countermeasures in CPU microarchitectures,

    J. Zhang, C. Chen, J. Cui, and K. Li, “Timing side-channel attacks and countermeasures in CPU microarchitectures,” ACM Computing Surveys, vol. 56, no. 7, pp. 1–40, 2024

    2024

  57. [65]

    Spy in the GPU-box: Covert and side channel attacks on multi-GPU systems,

    S. B. Dutta, H. Naghibijouybari, A. Gupta, N. Abu-Ghazaleh, A. Mar- quez, and K. Barker, “Spy in the GPU-box: Covert and side channel attacks on multi-GPU systems,” in Proceedings of the 50th Annual International Symposium on Computer Architecture , 2023, pp. 1–13

    2023

  58. [66]

    Intelligent mobile malware detection using permission requests and API calls,

    M. Alazab, M. Alazab, A. Shalaginov, A. Mesleh, and A. Awajan, “Intelligent mobile malware detection using permission requests and API calls,” Future Generation Computer Systems , vol. 107, pp. 509– 521, 2020

    2020

  59. [67]

    MAAR: Robust features to detect malicious activity based on API calls, their arguments and return values,

    Z. Salehi, A. Sami, and M. Ghiasi, “MAAR: Robust features to detect malicious activity based on API calls, their arguments and return values,” Engineering Applications of Artificial Intelligence , vol. 59, pp. 93–102, 2017

    2017

  60. [68]

    EnsembleHMD: Accurate hardware malware detec- tors with specialized ensemble classifiers,

    K. N. Khasawneh, M. Ozsoy, C. Donovick, N. Abu-Ghazaleh, and D. Ponomarev, “EnsembleHMD: Accurate hardware malware detec- tors with specialized ensemble classifiers,” IEEE Transactions on Dependable and Secure Computing, vol. 17, no. 3, pp. 620–633, 2018

    2018

  61. [69]

    Peeping Tom in the Neighborhood: Keystroke eavesdropping on multi-user systems

    K. Zhang and X. Wang, “Peeping Tom in the Neighborhood: Keystroke eavesdropping on multi-user systems.” in USENIX Security Symposium, vol. 20, 2009, p. 23

    2009

  62. [70]

    Collaborative TCP sequence number inference attack: how to crack sequence number under a second,

    Z. Qian, Z. M. Mao, and Y . Xie, “Collaborative TCP sequence number inference attack: how to crack sequence number under a second,” in Proceedings of the 2012 ACM conference on Computer and communications security , 2012, pp. 593–604

    2012

  63. [71]

    MES-attacks: Software-controlled covert channels based on mutual exclusion and synchronization,

    C. Shen, J. Zhang, and G. Qu, “MES-attacks: Software-controlled covert channels based on mutual exclusion and synchronization,” in 2023 60th ACM/IEEE Design Automation Conference (DAC). IEEE, 2023, pp. 1–6

    2023

  64. [72]

    Lockeddown: Exploiting contention on Host-GPU PCIe bus for fun and profit,

    M. Side, F. Yao, and Z. Zhang, “Lockeddown: Exploiting contention on Host-GPU PCIe bus for fun and profit,” in 2022 IEEE 7th European Symposium on Security and Privacy (EuroS&P) , 2022, pp. 270–285

    2022

  65. [73]

    Lord of the ring (s): Side channel attacks on the CPU On-Chip ring interconnect are practical,

    R. Paccagnella, L. Luo, and C. W. Fletcher, “Lord of the ring (s): Side channel attacks on the CPU On-Chip ring interconnect are practical,” in 30th USENIX Security Symposium (USENIX Security 21) , 2021, pp. 645–662

    2021

  66. [74]

    Leaky buddies: Cross-component covert channels on integrated CPU-GPU systems,

    S. B. Dutta, H. Naghibijouybari, N. Abu-Ghazaleh, A. Marquez, and K. Barker, “Leaky buddies: Cross-component covert channels on integrated CPU-GPU systems,” in 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA) . IEEE, 2021, pp. 972–984

    2021

  67. [75]

    MeshUp: Stateless cache side- channel attack on CPU mesh,

    J. Wan, Y . Bi, Z. Zhou, and Z. Li, “MeshUp: Stateless cache side- channel attack on CPU mesh,” in 2022 IEEE Symposium on Security and Privacy (SP) . IEEE, 2022, pp. 1506–1524

    2022

  68. [76]

    Page cache attacks,

    D. Gruss, E. Kraft, T. Tiwari, M. Schwarz, A. Trachtenberg, J. Hen- nessey, A. Ionescu, and A. Fogh, “Page cache attacks,” inProceedings of the 2019 ACM SIGSAC Conference on Computer and Communi- cations Security, 2019, pp. 167–180

    2019

  69. [77]

    Hey, you, get off of my cloud: exploring information leakage in third-party compute clouds,

    T. Ristenpart, E. Tromer, H. Shacham, and S. Savage, “Hey, you, get off of my cloud: exploring information leakage in third-party compute clouds,” in Proceedings of the 16th ACM conference on Computer and communications security, 2009, pp. 199–212

    2009

  70. [78]

    A study on the security implications of information leakages in container clouds,

    X. Gao, B. Steenkamer, Z. Gu, M. Kayaalp, D. Pendarakis, and H. Wang, “A study on the security implications of information leakages in container clouds,” IEEE Transactions on Dependable and Secure Computing, vol. 18, no. 1, pp. 174–191, 2018

    2018

  71. [79]

    Last- level cache side-channel attacks are feasible in the modern public cloud,

    Z. N. Zhao, A. Morrison, C. W. Fletcher, and J. Torrellas, “Last- level cache side-channel attacks are feasible in the modern public cloud,” in Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume...

    2024

  72. [80]

    Everywhere all at once: Co-location attacks on public cloud FaaS,

    ——, “Everywhere all at once: Co-location attacks on public cloud FaaS,” in Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 1, 2024, pp. 133–149. Appendix A. Meta-Review A.1. Summary This pape...

    2024

  73. [81]

    It demonstrates how to carefully control and leverage syncfs delays to instantiate covert channels and infer behaviors of victim pro- cesses

    This paper identifies an impactful vulnerability in file system code that is present in virtually all exist- ing Linux systems. It demonstrates how to carefully control and leverage syncfs delays to instantiate covert channels and infer behaviors of victim pro- cesses

  74. [82]

    Secure file and storage systems have been well-studied, but recent security research has shown that side-channel attacks are present at virtually all layers of the software stack

    The paper provides a valuable step forward in an established field. Secure file and storage systems have been well-studied, but recent security research has shown that side-channel attacks are present at virtually all layers of the software stack. This paper highlights that ev...

  75. [83]

    Attacks therefore have anomalous behavior that might be easily detected by simple (or state of the art) anomaly detection techniques

    The attacks rely on high sampling rates (150K calls per second). Attacks therefore have anomalous behavior that might be easily detected by simple (or state of the art) anomaly detection techniques. The takeaways could be stronger if the paper evaluated the detectability of th...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.