Deep Learning with Differential Privacy

Abadi, M., Chu, A., Goodfellow, I., McMahan, H. B., Mironov, I., Talwar, K., & Zhang, L. (2016). Deep Learning with Differential Privacy. Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security (CCS), 308-318.

Type

Conference paper

Intellectual base

Primary research

Year

2016

Pages

308-318

DOI

10.1145/2976749.2978318

DOI registrant

Association for Computing Machinery

Link

https://doi.org/10.1145/2976749.2978318

Cited from

computer_science

Cited by

1 citation across 1 artifact.

Each citation links to the sentence it supports in the citing article.

Primes

• Hidden Information Reconstruction
  • The role mappings are regular and substrate-spanning: the protected input maps to a key, a training record, a redacted word, a tissue density, a private intention; the output channel maps to a power trace, an encrypted packet, a vector embedding, a shared gradient, a CT projection; the prior maps to natural-image statistics, corpus statistics, physical models, expected-text distributions; the defence maps to DP-SGD, padding, decoy traffic, synthetic cohorts.

    This sourceIntroduces DP-SGD, moving differential-privacy noise into ML training as the channel-narrowing defence against gradient/membership leakage.

    Supported in partVerified against the work's full text

    Abadi et al. instantiate one role mapping only: training records as protected input, with DP-SGD as the defence; the cross-substrate regularity is not discussed.

    “compute the average, add noise in order to protect privacy”

    Read by an automated reader; not a human review. How support was checked

Verification

Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.

Does it back the claim? Read against the text for 1 of 1 citation: 1 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.

Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.

See how references were verified.

Registry ID ref:1cb32ad2ce79 · see in the full table