Papers › An Empirical Study on the Intrinsic Privacy of SGD

An Empirical Study on the Intrinsic Privacy of SGD

5 Dec 2019arXiv:1912.02919archive 2025-07-28

Stephanie L. Hyland, Shruti Tople

Introducing noise in the training of machine learning systems is a powerful way to protect individual privacy via differential privacy guarantees, but comes at a cost to utility. This work looks at whether the inherent randomness of stochastic gradient descent (SGD) could contribute to privacy, effectively reducing the amount of additional noise required to achieve a given privacy guarantee. We conduct a large-scale empirical study to examine this question. Training a grid of over 120,000 models across four datasets (tabular and images) on convex and non-convex objectives, we demonstrate that the random seed has a larger impact on model weights than any individual training example. We test the distribution over weights induced by the seed, finding that the simple convex case can be modelled with a multivariate Gaussian posterior, while neural networks exhibit multi-modal and non-Gaussian weight distributions. By casting convex SGD as a Gaussian mechanism, we then estimate an `intrinsic' data-dependent ϵᵢ(𝒟), finding values as low as 6.3, dropping to 1.9 using empirical estimates. We use a membership inference attack to estimate ϵ for non-convex SGD and demonstrate that hiding the random seed from the adversary results in a statistically significant reduction in attack performance, corresponding to a reduction in the effective ϵ. These results provide empirical evidence that SGD exhibits appreciable variability relative to its dataset sensitivity, and this `intrinsic noise' has the potential to be leveraged to improve the utility of privacy-preserving machine learning.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 16 code samples harvested from 1 repository linked to this paper: 11 ran (of which 0 constructed an object rather than computing a result; 11 with no instrument failure: 0 honoured, 0 violated, 11 with no contract checked; 0 where Syntology's instrument failed) · 5 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code; here there are none.

By repository: official repository: 16 samples from 1 repository, 11 ran (11 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

microsoft/intrinsic-private-sgd officialmentioned in papertfMITSyntology: 16 samples, 11 ran (0 instrument-failed) · 5 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

16 samples harvested: 11 ran (of which 0 constructed an object rather than computing a result; 11 with no instrument failure: 0 honoured, 0 violated, 11 with no contract checked; 0 where Syntology's instrument failed) · 5 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

11ran
5unverified

Licence: 0 of the 16 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from microsoft/intrinsic-private-sgd. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

microsoft/intrinsic-private-sgd official repository 16 samples, 11 ran (0 instrument-failed) · 5 unverified
analyse_mi_results aDPSGD/run_mi_attack.py ran MIT (permissive) · fe9612fdf18b362f · report fixture: signature df:DataFrame/0/? · returned DataFrame · level recorded 2026-09-25
compute_additional_noise aDPSGD/noise_utils.py ran fingerprinted MIT (permissive) · 26de49e46eb9bd78 · report fixture: signature target_sigma:ndarray/1/float, intrinsic_noise:ndarray/1/float · returned ndarray · level recorded 2026-09-25
compute_gaussian_noise aDPSGD/noise_utils.py ran MIT (permissive) · 7d426b2ef7951163 · report fixture: signature epsilon:float/0/float, delta:float/0/float, sensitivity:float/0/float, verbose:bool/0/bool · returned float64 · level recorded 2026-09-25
compute_wu_bound_strong aDPSGD/noise_utils.py ran MIT (permissive) · e24eb576415f7535 · report fixture: signature lipschitz_constant:float/0/float, gamma:float/0/float, n_samples:int/0/int, batch_size:int/0/int, verbose:bool/0/bool · returned float · level recorded 2026-09-25
get_classifier in 16 papers aDPSGD/attacks.py ran MIT (permissive) · a8aefe19dea792d8 · report fixture: signature train_act:ndarray/2/float, test_act:ndarray/2/float · returned RandomForestClassifier · level recorded 2026-09-25
get_dataset_size aDPSGD/experiment_metadata.py ran MIT (permissive) · ebe56f1b69048664 · report fixture: signature data_cfg:dict/0/? · returned int · level recorded 2026-09-25
get_model_init_path aDPSGD/cfg_utils.py ran MIT (permissive) · 3c737e63cc13a19a · report fixture: signature cfg:dict/0/?, diffinit:bool/0/bool · returned PosixPath · level recorded 2026-09-25
get_n_weights aDPSGD/experiment_metadata.py ran MIT (permissive) · 503264e42be535d8 · report fixture: signature cfg:dict/0/? · returned int · level recorded 2026-09-25
get_threshold in 7 papers aDPSGD/attacks.py ran MIT (permissive) · 8bce23044e7f4920 · report fixture: signature train_loss:list/1/float · returned float · level recorded 2026-09-25
min_max_rescale aDPSGD/data_utils.py ran MIT (permissive) · 649b482fe8ad62e8 · report fixture: signature df_train:ndarray/2/float, df_test:ndarray/2/float, good_columns:NoneType/0/? · returned tuple · level recorded 2026-09-25
validation_split in 5 papers aDPSGD/data_utils.py ran MIT (permissive) · 2c6c3a9b821d4b3a · report fixture: signature x_train:ndarray/2/float, y_train:ndarray/1/int, x_test:ndarray/2/float, y_test:ndarray/1/int, replace_index:int/0/int · returned tuple · level recorded 2026-09-25
build_model in 469 papers aDPSGD/model_utils.py unverified MIT (permissive) · aa8df6e3e5373e58 · report
define_metric_functions aDPSGD/model_utils.py unverified MIT (permissive) · ae0a33bbf1d9dc11 · report
get_mi_attack_accuracy aDPSGD/attacks.py unverified MIT (permissive) · 2658ac28c563e3a6 · report
load_cfg in 17 papers aDPSGD/cfg_utils.py unverified MIT (permissive) · af89613292776213 · report
load_model_at_time aDPSGD/model_utils.py unverified MIT (permissive) · e922c575ce82dfaa · report

Tasks

Inference AttackMembership Inference AttackPrivacy Preserving

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

SGDTest

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections