From 60fc99b7ef13809da7fa3cdd42144353dec67c0d Mon Sep 17 00:00:00 2001 From: Chuck McCallum Date: Tue, 23 Jun 2026 11:29:15 -0400 Subject: [PATCH] rename "preprocessing" --- deployments/apple_iconic_scenes_photo_selections_2023.yaml | 2 +- deployments/apple_popular_emojis.yaml | 2 +- deployments/assistive_ai.yaml | 2 +- deployments/audience_engagement.yaml | 2 +- deployments/autoplay_intent.yaml | 2 +- deployments/birth_dataset.yaml | 2 +- deployments/broadband_coverage.yaml | 2 +- deployments/census_demographic_and_housing.yaml | 2 +- deployments/census_redistricting_data.yaml | 2 +- deployments/county_business_patterns.yaml | 2 +- deployments/covid19_notifications.yaml | 2 +- deployments/covid19_search_trends_symptoms.yaml | 2 +- deployments/employment_outcomes.yaml | 2 +- deployments/google_mobility.yaml | 2 +- deployments/healthkit.yaml | 2 +- deployments/korean_statistics_datahub.yaml | 2 +- deployments/linkedin_hiring_reports.yaml | 2 +- deployments/lookup_hints.yaml | 2 +- ...crosoft_ctdc_victim_perpertartor_synthetic_dataset_2023.yaml | 2 +- deployments/mobility_trends_hurricane.yaml | 2 +- deployments/movement_ranges_maps.yaml | 2 +- deployments/on_device_browser_rec.yaml | 2 +- deployments/private_third_party_audits.yaml | 2 +- deployments/recurve_energy_dp.yaml | 2 +- deployments/safari_energy_drain.yaml | 2 +- deployments/safety_classifier.yaml | 2 +- deployments/sas_data_maker_vulnerable_persons.yaml | 2 +- deployments/shared_mobility_dataset.yaml | 2 +- deployments/spanish_language_next_word.yaml | 2 +- deployments/synth_data_public_use.yaml | 2 +- deployments/telemetry_collection_windows.yaml | 2 +- deployments/uber.yaml | 2 +- deployments/user_url_privacy.yaml | 2 +- deployments/vaccine_search_insights.yaml | 2 +- deployments/wikimedia_current_usage_data.yaml | 2 +- deployments/wikimedia_editor_activity_statistics_2023.yaml | 2 +- deployments/wikimedia_historical_usage_data.yaml | 2 +- .../wikimedia_russian_editor_activity_statistics_2023.yaml | 2 +- schemas/deployments-schema.yaml | 2 +- tests/good_deployments/google_tier_3.yaml | 2 +- tests/good_deployments/template.yaml | 2 +- ui-hints/deployments-hints.yaml | 2 +- 42 files changed, 42 insertions(+), 42 deletions(-) diff --git a/deployments/apple_iconic_scenes_photo_selections_2023.yaml b/deployments/apple_iconic_scenes_photo_selections_2023.yaml index 9df72aa..14cebcf 100644 --- a/deployments/apple_iconic_scenes_photo_selections_2023.yaml +++ b/deployments/apple_iconic_scenes_photo_selections_2023.yaml @@ -48,7 +48,7 @@ deployment: # composition: '' # Tier 3 implementation: - # pre_processing_eda_hyperparameter_tuning: '' # Tier 3 + # preprocessing_and_hyperparameter_tuning: '' # Tier 3 mechanisms: | - Randomized Response with amplification through the secure aggregation protocol. The local Differential Privacy mechanism encodes a single, randomly sampled location-category pair into a tiled one-hot vector, subsequently applying randomized bit-flipping to satisfy DP guarantees. - No explicit information regarding the privacy unit was provided on the posted online article by Apple: https://machinelearning.apple.com/research/scenes-differential-privacy diff --git a/deployments/apple_popular_emojis.yaml b/deployments/apple_popular_emojis.yaml index d6a4d54..0867cd2 100644 --- a/deployments/apple_popular_emojis.yaml +++ b/deployments/apple_popular_emojis.yaml @@ -59,7 +59,7 @@ deployment: composition: No information was provided about composition. implementation: - pre_processing_eda_hyperparameter_tuning: 'Three parameters are explicitly set: \\(\epsilon = 4\\), number of hash functions k = 65,546 and the size of the privatized vector m = 1024. These were predetermined by the authors based on an analysis of the tradeoff between utility, server computation cost (related to k), and device bandwidth (related to m).' + preprocessing_and_hyperparameter_tuning: 'Three parameters are explicitly set: \\(\epsilon = 4\\), number of hash functions k = 65,546 and the size of the privatized vector m = 1024. These were predetermined by the authors based on an analysis of the tradeoff between utility, server computation cost (related to k), and device bandwidth (related to m).' mechanisms: | - The differential privacy algorithm used is Private Count Mean Sketch (CMS). - No information was provided about the implementation of the mechanism. diff --git a/deployments/assistive_ai.yaml b/deployments/assistive_ai.yaml index d7b3fc3..9804293 100644 --- a/deployments/assistive_ai.yaml +++ b/deployments/assistive_ai.yaml @@ -27,7 +27,7 @@ deployment: model: model_name: Central implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Differentially Private Set Union justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/audience_engagement.yaml b/deployments/audience_engagement.yaml index 83f42ee..f0e3db4 100644 --- a/deployments/audience_engagement.yaml +++ b/deployments/audience_engagement.yaml @@ -31,7 +31,7 @@ deployment: release_type_description: The data are refreshed on a monthly basis. The budget also resets on a monthly basis. access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace, Gumbel justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/autoplay_intent.yaml b/deployments/autoplay_intent.yaml index c1b3351..4afd0c1 100644 --- a/deployments/autoplay_intent.yaml +++ b/deployments/autoplay_intent.yaml @@ -29,7 +29,7 @@ deployment: release_type_description: Daily access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Private Count Mean Sketch justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/birth_dataset.yaml b/deployments/birth_dataset.yaml index d747a59..a495f39 100644 --- a/deployments/birth_dataset.yaml +++ b/deployments/birth_dataset.yaml @@ -31,7 +31,7 @@ deployment: release_type_description: Static access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: PrivBayes with Private Selection (Universal Microdata Scheme) justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/broadband_coverage.yaml b/deployments/broadband_coverage.yaml index a3a2527..96d7757 100644 --- a/deployments/broadband_coverage.yaml +++ b/deployments/broadband_coverage.yaml @@ -64,7 +64,7 @@ deployment: By sequential composition, the total privacy loss is 0.2. (Note: The equation for Broadband Coverage Estimate does not use the "number of devices with speed < 25". In fact, it seems this number is never used.) implementation: - pre_processing_eda_hyperparameter_tuning: The 32,653 zip codes were predefined based on the average number of machines in the last years (not on the dataset being protected). + preprocessing_and_hyperparameter_tuning: The 32,653 zip codes were predefined based on the average number of machines in the last years (not on the dataset being protected). mechanisms: | - The Laplace mechanism with \\(\epsilon = 0.1\\) was applied to every count query (see section on Composition for query descriptions). diff --git a/deployments/census_demographic_and_housing.yaml b/deployments/census_demographic_and_housing.yaml index 877d46f..cb7ee63 100644 --- a/deployments/census_demographic_and_housing.yaml +++ b/deployments/census_demographic_and_housing.yaml @@ -76,7 +76,7 @@ deployment: - For the TopDownAlgorithm, the total \\(\rho\\) is the sum of the \\(\rho\\) values allocated to all queries at every level of the geographic hierarchy (nation, state, county, tract, block group, and block). implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - Preprocessing: the raw census responses undergo pre-processing steps to create the CEF. This involves: - Editing: Ensuring logical consistencies among characteristics for a person or household. For instance, an edit constraint ensures that a mother must be older than her natural child. - Imputation: Filling in missing or misreported data for key characteristics like age, sex, race, and ethnicity to ensure every record is complete. diff --git a/deployments/census_redistricting_data.yaml b/deployments/census_redistricting_data.yaml index fdf616c..46ddb80 100644 --- a/deployments/census_redistricting_data.yaml +++ b/deployments/census_redistricting_data.yaml @@ -76,7 +76,7 @@ deployment: - For the TopDownAlgorith, the total \\(\rho\\) is the sum of the \\(\rho\\) values allocated to all queries at every level of the geographic hierarchy (nation, state, county, tract, block group, and block). implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - Preprocessing: the raw census responses undergo pre-processing steps to create the CEF. This involves: - Editing: Ensuring logical consistencies among characteristics for a person or household. For instance, an edit constraint ensures that a mother must be older than her natural child. - Imputation: Filling in missing or misreported data for key characteristics like age, sex, race, and ethnicity to ensure every record is complete. diff --git a/deployments/county_business_patterns.yaml b/deployments/county_business_patterns.yaml index 217d200..a7f7baa 100644 --- a/deployments/county_business_patterns.yaml +++ b/deployments/county_business_patterns.yaml @@ -29,7 +29,7 @@ deployment: release_type_description: Static access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Gaussian justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/covid19_notifications.yaml b/deployments/covid19_notifications.yaml index 1e87362..0ef8bff 100644 --- a/deployments/covid19_notifications.yaml +++ b/deployments/covid19_notifications.yaml @@ -31,7 +31,7 @@ deployment: release_type_description: Varies access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Randomized Response justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/covid19_search_trends_symptoms.yaml b/deployments/covid19_search_trends_symptoms.yaml index ea3b05c..7d4500e 100644 --- a/deployments/covid19_search_trends_symptoms.yaml +++ b/deployments/covid19_search_trends_symptoms.yaml @@ -72,7 +72,7 @@ deployment: - Level 2 (e.g., County): \\(\epsilon_2'\\) = 0.014 implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - Individual search queries are mapped to one or more symptoms from a predefined list. - The mapped queries are aggregated by user, day, symptom, and geographic region to produce raw counts. - For each day and geographic granularity level, a user can contribute at most once to any given count (per-symptom bound) and to no more than three counts in total (cross-symptom bound). Any symptom searches beyond this limit are discarded. diff --git a/deployments/employment_outcomes.yaml b/deployments/employment_outcomes.yaml index f1ed4b5..96da90e 100644 --- a/deployments/employment_outcomes.yaml +++ b/deployments/employment_outcomes.yaml @@ -44,7 +44,7 @@ deployment: # composition: # implementation: - # pre_processing_eda_hyperparameter_tuning: + # preprocessing_and_hyperparameter_tuning: # mechanisms: # justification: resources: diff --git a/deployments/google_mobility.yaml b/deployments/google_mobility.yaml index b076f62..2eca046 100644 --- a/deployments/google_mobility.yaml +++ b/deployments/google_mobility.yaml @@ -80,7 +80,7 @@ deployment: - Composition: A user contributes at most once per geographic level. The total budget is the sum across the levels: 0.11*2 + 0.22 = 0.44 implementation: - pre_processing_eda_hyperparameter_tuning: 'User contributions for public places are bounded: If a user visits more than four (category, location) pairs on a given day at a specific geographic level, four are chosen at random, and the rest are discarded. Each user can also contribute at most once per category at a given location.' + preprocessing_and_hyperparameter_tuning: 'User contributions for public places are bounded: If a user visits more than four (category, location) pairs on a given day at a specific geographic level, four are chosen at random, and the rest are discarded. Each user can also contribute at most once per category at a given location.' mechanisms: | - The Laplace mechanism was used for simple counts, like the number of daily visits to public places and workplaces. The DP Mean Mechanism: used to calculate the average time spent at residences. - The mechanisms were implemented via Google's open-source C++ Differential Privacy Library: diff --git a/deployments/healthkit.yaml b/deployments/healthkit.yaml index 3163602..5573b44 100644 --- a/deployments/healthkit.yaml +++ b/deployments/healthkit.yaml @@ -32,7 +32,7 @@ deployment: implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Private Count Mean Sketch justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/korean_statistics_datahub.yaml b/deployments/korean_statistics_datahub.yaml index 677c3ef..84caa6c 100644 --- a/deployments/korean_statistics_datahub.yaml +++ b/deployments/korean_statistics_datahub.yaml @@ -27,7 +27,7 @@ deployment: release_type_description: Not specified access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Unknown justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/linkedin_hiring_reports.yaml b/deployments/linkedin_hiring_reports.yaml index b400bee..475c6a7 100644 --- a/deployments/linkedin_hiring_reports.yaml +++ b/deployments/linkedin_hiring_reports.yaml @@ -35,7 +35,7 @@ deployment: release_type_description: Monthly access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace, Gumbel justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/lookup_hints.yaml b/deployments/lookup_hints.yaml index ff64312..571d522 100644 --- a/deployments/lookup_hints.yaml +++ b/deployments/lookup_hints.yaml @@ -32,7 +32,7 @@ deployment: release_type_description: Daily access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Count Mean Sketch or Hadamard Count Mean Sketch justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/microsoft_ctdc_victim_perpertartor_synthetic_dataset_2023.yaml b/deployments/microsoft_ctdc_victim_perpertartor_synthetic_dataset_2023.yaml index c55cd90..4c09ead 100644 --- a/deployments/microsoft_ctdc_victim_perpertartor_synthetic_dataset_2023.yaml +++ b/deployments/microsoft_ctdc_victim_perpertartor_synthetic_dataset_2023.yaml @@ -63,7 +63,7 @@ deployment: # composition: '' # Tier 3 implementation: - # pre_processing_eda_hyperparameter_tuning: '' # Tier 3 + # preprocessing_and_hyperparameter_tuning: '' # Tier 3 mechanisms: | - Continuous Gaussian mechanism - The dataset was generated using Microsoft's Synthetic Data Showcase, which follows a two-stage approach: diff --git a/deployments/mobility_trends_hurricane.yaml b/deployments/mobility_trends_hurricane.yaml index 035f970..3b8d864 100644 --- a/deployments/mobility_trends_hurricane.yaml +++ b/deployments/mobility_trends_hurricane.yaml @@ -29,7 +29,7 @@ deployment: release_type_description: Not specified access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/movement_ranges_maps.yaml b/deployments/movement_ranges_maps.yaml index 8116b23..0f8fd4c 100644 --- a/deployments/movement_ranges_maps.yaml +++ b/deployments/movement_ranges_maps.yaml @@ -28,7 +28,7 @@ deployment: data_source_type: Dynamic access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/on_device_browser_rec.yaml b/deployments/on_device_browser_rec.yaml index d3f67c2..bf0b58c 100644 --- a/deployments/on_device_browser_rec.yaml +++ b/deployments/on_device_browser_rec.yaml @@ -25,7 +25,7 @@ deployment: release_type_description: Not specified access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: DP SGD* justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/private_third_party_audits.yaml b/deployments/private_third_party_audits.yaml index 1f5338f..d1acb0b 100644 --- a/deployments/private_third_party_audits.yaml +++ b/deployments/private_third_party_audits.yaml @@ -30,7 +30,7 @@ deployment: release_type_description: Static access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Not Specificed, Synthetic Data justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/recurve_energy_dp.yaml b/deployments/recurve_energy_dp.yaml index b3bd96f..b7788cb 100644 --- a/deployments/recurve_energy_dp.yaml +++ b/deployments/recurve_energy_dp.yaml @@ -75,7 +75,7 @@ deployment: - Because the comparison group was a sample of a larger population, the paper applies an 'amplification factor' of 0.124 to reduce the total privacy cost. The correct amplification formula yields \\(\epsilon = 4.72\\) and \\(\delta = 5.06e-9\\). implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - The CalTRACK methodology was used to model and generate a 'predicted' (counterfactual) hourly energy usage for every customer in both the treatment (non-sensitive) and comparison (sensitive) groups. - An upper bound was selected for hourly consumption data used for clipping in a two-step process: 1. Build a histogram using \\(\epsilon = 0.1\\) to understand the distribution of the data. This private histogram was then used to generate a synthetic dataset, which was used to choose the privacy budget \\(epsilon=0.2\\) to be used for selecting clamping bound using the Sparse Vector Technique. diff --git a/deployments/safari_energy_drain.yaml b/deployments/safari_energy_drain.yaml index 56b5dbf..37c54a2 100644 --- a/deployments/safari_energy_drain.yaml +++ b/deployments/safari_energy_drain.yaml @@ -45,7 +45,7 @@ deployment: composition: No information was provided about composition. implementation: - pre_processing_eda_hyperparameter_tuning: 'No hyperperameters are required; no EDA is described' + preprocessing_and_hyperparameter_tuning: 'No hyperperameters are required; no EDA is described' mechanisms: Private Hadamard Count Mean Sketch (PHCMS) justification: '"Formal proof, described in Theorem 4.1, Theorem 4.3"' diff --git a/deployments/safety_classifier.yaml b/deployments/safety_classifier.yaml index 1316d19..446a995 100644 --- a/deployments/safety_classifier.yaml +++ b/deployments/safety_classifier.yaml @@ -42,7 +42,7 @@ deployment: # composition: # implementation: - # pre_processing_eda_hyperparameter_tuning: + # preprocessing_and_hyperparameter_tuning: # mechanisms: # justification: resources: diff --git a/deployments/sas_data_maker_vulnerable_persons.yaml b/deployments/sas_data_maker_vulnerable_persons.yaml index d9b3c18..46b49d4 100644 --- a/deployments/sas_data_maker_vulnerable_persons.yaml +++ b/deployments/sas_data_maker_vulnerable_persons.yaml @@ -50,7 +50,7 @@ deployment: - Group privacy: To account for the one-to-many relationship between Customers and Accounts, the number of accounts per customer is bounded to 5 in the second PrivBayes network model, ensuring privacy loss accounting under group privacy. implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - Preprocessing: | - discarding PIIs like real account/transaction/merchant IDs, and merchant names; generating random synthetic IDs/names in a similar format - discretizing numerical data in the Customers and Accounts tables (required for PrivBayes models, which operate on discrete/categorical data); after generation, values are transformed back to numerical by randomly sampling from the discretized bins diff --git a/deployments/shared_mobility_dataset.yaml b/deployments/shared_mobility_dataset.yaml index 81a9528..645677e 100644 --- a/deployments/shared_mobility_dataset.yaml +++ b/deployments/shared_mobility_dataset.yaml @@ -34,7 +34,7 @@ deployment: release_type_description: Static access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace Mechanism justification: '"All trips were anonymized and aggregated by jointly applying differential privacy via the Laplace mechanism in combination with k-anonymity."' resources: diff --git a/deployments/spanish_language_next_word.yaml b/deployments/spanish_language_next_word.yaml index 391bbcd..f4da038 100644 --- a/deployments/spanish_language_next_word.yaml +++ b/deployments/spanish_language_next_word.yaml @@ -53,7 +53,7 @@ deployment: composition: The article does not specifically mention composition, only that the final privacy guarantee is rho = 0.81 zCDP. It mentions that training ran for 2000 rounds over six days, and devices participated in training at most once every 24 hours. implementation: - pre_processing_eda_hyperparameter_tuning: No information is provided about pre-processing steps, EDA, or the process by which hyperparameters (privacy loss parameter or number of training rounds) were tuned. + preprocessing_and_hyperparameter_tuning: No information is provided about pre-processing steps, EDA, or the process by which hyperparameters (privacy loss parameter or number of training rounds) were tuned. mechanisms: | - The mechanism is a DP variant of Follow-The-Regularized-Leader (DP-FTRL), which utilizes negatively correlated noise, added by the aggregating server (essentially, adding noise to one gradient and subtracting that same noise from a later gradient) and efficiently implements this using the Tree Aggregation algorithm. - The mechanism was implemented using Google’s open-source libraries TenserFlow Federated and TensorFlow Privacy. diff --git a/deployments/synth_data_public_use.yaml b/deployments/synth_data_public_use.yaml index f344838..4d5c58b 100644 --- a/deployments/synth_data_public_use.yaml +++ b/deployments/synth_data_public_use.yaml @@ -26,7 +26,7 @@ deployment: release_type_description: Static access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: DP GAN* justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/telemetry_collection_windows.yaml b/deployments/telemetry_collection_windows.yaml index 78bb4b7..50775de 100644 --- a/deployments/telemetry_collection_windows.yaml +++ b/deployments/telemetry_collection_windows.yaml @@ -30,7 +30,7 @@ deployment: data_source_type: Dynamic access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace, Randomized Rounding justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/uber.yaml b/deployments/uber.yaml index deab5c0..9d5b1ec 100644 --- a/deployments/uber.yaml +++ b/deployments/uber.yaml @@ -53,7 +53,7 @@ deployment: - deployment does not manage privacy budgets - user can compute overall privacy loss using strong composition bounds on sequential queries implementation: - pre_processing_eda_hyperparameter_tuning: none described + preprocessing_and_hyperparameter_tuning: none described mechanisms: | - The differential privacy algorithm used is bounded epsilon-delta DP - details: diff --git a/deployments/user_url_privacy.yaml b/deployments/user_url_privacy.yaml index 7b108f2..a421547 100644 --- a/deployments/user_url_privacy.yaml +++ b/deployments/user_url_privacy.yaml @@ -39,7 +39,7 @@ deployment: release_type_description: Static access_type: Non-interactive implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Laplace, Modified Report Noisy Max justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/vaccine_search_insights.yaml b/deployments/vaccine_search_insights.yaml index f70b1b0..6c8d703 100644 --- a/deployments/vaccine_search_insights.yaml +++ b/deployments/vaccine_search_insights.yaml @@ -31,7 +31,7 @@ deployment: release_type_description: Daily access_type: Non-interactive # TODO: Is this correct? implementation: - pre_processing_eda_hyperparameter_tuning: '' # TODO: Fill in correct value + preprocessing_and_hyperparameter_tuning: '' # TODO: Fill in correct value mechanisms: Gaussian justification: '' # TODO: Fill in correct value resources: diff --git a/deployments/wikimedia_current_usage_data.yaml b/deployments/wikimedia_current_usage_data.yaml index 8554811..d92a36f 100644 --- a/deployments/wikimedia_current_usage_data.yaml +++ b/deployments/wikimedia_current_usage_data.yaml @@ -53,7 +53,7 @@ deployment: composition: 'There is no need for composition analysis: The data is partitioned by day, and there is no overlap between days, so the privacy loss does not accumulate.' implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - Client-side filtering: Before pageview data is sent to the server, a client-side algorithm annotates each contribution with a boolean flag, which indicates whether the contribution should be included in the server side computation, the criteria being that only the first 10 unique pageviews per user, per day are included. The client uses a salted hash of the page ID to track which pages have already been counted, rather than storing raw identifiers, further reducing fingerprinting risk. This pre-processing step enables user-privacy without sending a user identifier to the server (which would be cookie-based and consistent across days). - The server determines the list of groups for which to release statistics by identifying all pages that have more than a global pageview threshold (which was chosen based on the true data with no DP) and create the cross-product of those pages with a pre-defined list of countries. - The hyperparameters per-user daily contribution bound, ingestion threshold, and suppression threshold (see “Unprotected Quantities” for details) were selected by computing metrics using the true data. The authors acknowledge that the parameters are not differentially private. diff --git a/deployments/wikimedia_editor_activity_statistics_2023.yaml b/deployments/wikimedia_editor_activity_statistics_2023.yaml index 07afa38..650598e 100644 --- a/deployments/wikimedia_editor_activity_statistics_2023.yaml +++ b/deployments/wikimedia_editor_activity_statistics_2023.yaml @@ -48,7 +48,7 @@ deployment: # composition: '' # Tier 3 implementation: - # pre_processing_eda_hyperparameter_tuning: '' # Tier 3 + # preprocessing_and_hyperparameter_tuning: '' # Tier 3 mechanisms: | - Laplace Mechanism - The processing pipelines consists of the following steps: diff --git a/deployments/wikimedia_historical_usage_data.yaml b/deployments/wikimedia_historical_usage_data.yaml index ac5fc17..8a640bc 100644 --- a/deployments/wikimedia_historical_usage_data.yaml +++ b/deployments/wikimedia_historical_usage_data.yaml @@ -63,7 +63,7 @@ deployment: composition: 'The results are published once with \\(\epsilon=1\\), and there is no additional privacy consumption.' implementation: - pre_processing_eda_hyperparameter_tuning: | + preprocessing_and_hyperparameter_tuning: | - The server determines the list of groups for which to release statistics by identifying all pages that have more than a global pageview threshold (which was chosen based on the true data with no DP) and create the cross-product of those pages with a pre-defined list of countries. - The hyperparameters per-user daily contribution bound, ingestion threshold, and suppression threshold (see “Unprotected Quantities” for details) were selected by computing metrics using the true data. The authors acknowledge that the parameters are not differentially private. mechanisms: | diff --git a/deployments/wikimedia_russian_editor_activity_statistics_2023.yaml b/deployments/wikimedia_russian_editor_activity_statistics_2023.yaml index c7dd374..733d145 100644 --- a/deployments/wikimedia_russian_editor_activity_statistics_2023.yaml +++ b/deployments/wikimedia_russian_editor_activity_statistics_2023.yaml @@ -48,7 +48,7 @@ deployment: # composition: '' # Tier 3 implementation: - # pre_processing_eda_hyperparameter_tuning: '' # Tier 3 + # preprocessing_and_hyperparameter_tuning: '' # Tier 3 mechanisms: | - Laplace Mechanism - The processing pipelines consists of the following steps: diff --git a/schemas/deployments-schema.yaml b/schemas/deployments-schema.yaml index d7dc366..ad1dba1 100644 --- a/schemas/deployments-schema.yaml +++ b/schemas/deployments-schema.yaml @@ -339,7 +339,7 @@ properties: type: object additionalProperties: False properties: - pre_processing_eda_hyperparameter_tuning: + preprocessing_and_hyperparameter_tuning: tier: 3 description: Description of any pre-processing of the data before DP protections and any exploratory data analysis conducted before DP protections, and whether privacy loss was accounted for. How were hyperparameters, like privacy loss parameters, tuned? Was privacy loss associated with this tuning accounted for? description_long: | diff --git a/tests/good_deployments/google_tier_3.yaml b/tests/good_deployments/google_tier_3.yaml index 6914cac..5735a94 100644 --- a/tests/good_deployments/google_tier_3.yaml +++ b/tests/good_deployments/google_tier_3.yaml @@ -41,7 +41,7 @@ deployment: Basic sequential composition (counting symptoms + normalization) implementation: - pre_processing_eda_hyperparameter_tuning: Not available + preprocessing_and_hyperparameter_tuning: Not available mechanisms: Laplace mechanism for both symptom search and normalization counts via Google’s open-source DP library justification: Not available resources: diff --git a/tests/good_deployments/template.yaml b/tests/good_deployments/template.yaml index f36fccd..c392667 100644 --- a/tests/good_deployments/template.yaml +++ b/tests/good_deployments/template.yaml @@ -49,7 +49,7 @@ deployment: composition: '' # Tier 3 implementation: - pre_processing_eda_hyperparameter_tuning: '' # Tier 3 + preprocessing_and_hyperparameter_tuning: '' # Tier 3 mechanisms: '' # Tier 3 justification: '' # Tier 3 diff --git a/ui-hints/deployments-hints.yaml b/ui-hints/deployments-hints.yaml index 66114c9..0e6467c 100644 --- a/ui-hints/deployments-hints.yaml +++ b/ui-hints/deployments-hints.yaml @@ -20,4 +20,4 @@ extra_columns: "Model": /deployment/model/model_name tile_names: # These tiles in the table need shorter names. - pre_processing_eda_hyperparameter_tuning: Pre Processing + preprocessing_and_hyperparameter_tuning: Pre Processing