1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236

use crate::errors::*;

use crate::{proto, Warnable, base};

use crate::components::{Component, Sensitivity, Expandable};
use crate::base::{IndexKey, Value, NodeProperties, AggregatorProperties, SensitivitySpace, ValueProperties, DataType, NatureContinuous, Nature, Vector1DNull, Jagged};
use crate::utilities::{prepend, get_literal};
use ndarray::{arr1, Array};
use indexmap::map::IndexMap;
use crate::utilities::inference::infer_property;


impl Component for proto::Histogram {
    fn propagate_property(
        &self,
        _privacy_definition: &Option<proto::PrivacyDefinition>,
        _public_arguments: IndexMap<base::IndexKey, &Value>,
        properties: NodeProperties,
        node_id: u32
    ) -> Result<Warnable<ValueProperties>> {
        let mut data_property = properties.get::<base::IndexKey>(&"data".into())
            .ok_or("data: missing")?.array()
            .map_err(prepend("data:"))?.clone();

        if !data_property.releasable {
            data_property.assert_is_not_aggregated()?;
        }

        // this check is already guaranteed by the state space, but still included for safety
        if data_property.data_type == DataType::Unknown {
            return Err("data_type must be known".into())
        }

        let categories = data_property.categories()?;

        if categories.num_columns() != 1 {
            return Err("data must contain one column".into())
        }
        let num_columns = data_property.num_columns()?;

        // save a snapshot of the state when aggregating
        data_property.aggregator = Some(AggregatorProperties::new(
            proto::component::Variant::Histogram(self.clone()),
            properties, num_columns));

        data_property.nature = Some(Nature::Continuous(NatureContinuous {
            lower: Vector1DNull::Int((0..num_columns).map(|_| Some(0)).collect()),
            upper: Vector1DNull::Int((0..num_columns).map(|_| data_property.num_records.clone()).collect()),
        }));
        data_property.data_type = DataType::Int;
        data_property.num_records = Some(categories.num_records()[0] as i64);
        data_property.dataset_id = Some(node_id as i64);

        Ok(ValueProperties::Array(data_property).into())
    }
}


impl Expandable for proto::Histogram {
    /// If min and max are not supplied, but are known statically, then add them automatically
    /// Add nodes for clamp or digitize if categories or edges are passed
    fn expand_component(
        &self,
        _privacy_definition: &Option<proto::PrivacyDefinition>,
        component: &proto::Component,
        _public_arguments: &IndexMap<IndexKey, &Value>,
        properties: &NodeProperties,
        component_id: u32,
        mut maximum_id: u32,
    ) -> Result<base::ComponentExpansion> {

        let mut expansion = base::ComponentExpansion::default();

        let data_id = component.arguments().get::<IndexKey>(&"data".into())
            .ok_or_else(|| Error::from("data is a required argument to Histogram"))?.to_owned();

        let mut component = component.clone();

        match (
            component.arguments().get::<IndexKey>(&"edges".into()),
            component.arguments().get::<IndexKey>(&"categories".into())) {

            (Some(edges_id), None) => {
                // digitize
                let mut arguments = indexmap![
                    IndexKey::from("data") => data_id,
                    "edges".into() => *edges_id
                ];

                let prior_arguments = component.arguments();
                prior_arguments.get::<IndexKey>(&"null_value".into())
                    .map(|v| arguments.insert("null_value".into(), *v));
                prior_arguments.get::<IndexKey>(&"inclusive_left".into())
                    .map(|v| arguments.insert("inclusive_left".into(), *v));

                maximum_id += 1;
                let id_digitize = maximum_id;
                expansion.computation_graph.insert(id_digitize, proto::Component {
                    arguments: Some(proto::ArgumentNodeIds::new(arguments)),
                    variant: Some(proto::component::Variant::Digitize(proto::Digitize {})),
                    omit: true,
                    submission: component.submission,
                });
                component.arguments = Some(proto::ArgumentNodeIds::new(indexmap!["data".into() => id_digitize]));
                expansion.traversal.push(id_digitize);
            }

            (None, Some(categories_id)) => {
                // clamp
                let prior_arguments = component.arguments();
                let null_id = prior_arguments.get::<IndexKey>(&"null_value".into())
                    .ok_or_else(|| Error::from("null_value is a required argument to Histogram when categories are not known"))?;
                maximum_id += 1;
                let id_clamp = maximum_id;
                expansion.computation_graph.insert(id_clamp, proto::Component {
                    arguments: Some(proto::ArgumentNodeIds::new(indexmap![
                        "data".into() => data_id,
                        "categories".into() => *categories_id,
                        "null_value".into() => *null_id
                    ])),
                    variant: Some(proto::component::Variant::Clamp(proto::Clamp {})),
                    omit: true,
                    submission: component.submission,
                });
                component.arguments = Some(proto::ArgumentNodeIds::new(indexmap!["data".into() => id_clamp]));
                expansion.traversal.push(id_clamp);
            }

            (None, None) => {
                let data_property = properties.get::<IndexKey>(&"data".into())
                    .ok_or("data: missing")?.array()
                    .map_err(prepend("data:"))?.clone();

                if data_property.categories().is_err() {
                    return Err("either edges or categories must be supplied".into())
                }

                maximum_id += 1;
                let id_categories = maximum_id;
                let categories = properties.get::<IndexKey>(&"data".into())
                    .ok_or("data: missing")?.array()?.categories()?;
                let value = match categories {
                    Jagged::Int(jagged) => arr1(&jagged[0]).into_dyn().into(),
                    Jagged::Float(jagged) => arr1(&jagged[0]).into_dyn().into(),
                    Jagged::Bool(jagged) => arr1(&jagged[0]).into_dyn().into(),
                    Jagged::Str(jagged) => arr1(&jagged[0]).into_dyn().into(),
                };
                let (patch_node, categories_release) = get_literal(value, component.submission)?;
                expansion.computation_graph.insert(id_categories, patch_node);
                expansion.properties.insert(id_categories, infer_property(&categories_release.value, None, id_categories)?);
                expansion.releases.insert(id_categories, categories_release);
                component.insert_argument(&"categories".into(), id_categories);
            }

            (Some(_), Some(_)) => return Err("either edges or categories must be supplied".into())
        }

        expansion.computation_graph.insert(component_id, component);

        Ok(expansion)
    }
}


impl Sensitivity for proto::Histogram {
    /// Histogram sensitivities [are backed by the the proofs here](https://github.com/opendp/smartnoise-core/blob/master/whitepapers/sensitivities/counts/counts.pdf).
    fn compute_sensitivity(
        &self,
        privacy_definition: &proto::PrivacyDefinition,
        properties: &NodeProperties,
        sensitivity_type: &SensitivitySpace
    ) -> Result<Value> {
        let data_property = properties.get::<base::IndexKey>(&"data".into())
            .ok_or("data: missing")?.array()
            .map_err(prepend("data:"))?.clone();

        data_property.assert_is_not_aggregated()?;

        match sensitivity_type {
            SensitivitySpace::KNorm(k) => {

                use proto::privacy_definition::Neighboring;
                use proto::privacy_definition::Neighboring::{Substitute, AddRemove};
                let neighboring_type = Neighboring::from_i32(privacy_definition.neighboring)
                    .ok_or_else(|| Error::from("neighboring definition must be either \"AddRemove\" or \"Substitute\""))?;

                // when categories are defined, a disjoint group by query is performed
                let categories_length = data_property.categories()?.num_records()[0];

                let num_records = data_property.num_records;

                let num_columns = data_property.num_columns()?;

                macro_rules! wrap {
                    ($sensitivity:expr) => {
                        Ok(Array::from_shape_vec(
                            vec![categories_length as usize, num_columns as usize],
                            (0..categories_length)
                                .map(|_| (0..num_columns)
                                    .map(|_| $sensitivity)
                                    .collect::<Vec<_>>())
                                .flatten()
                                .collect::<Vec<_>>())?.into())
                    }
                }

                // SENSITIVITY DERIVATIONS
                match (neighboring_type, categories_length, num_records) {
                    // one category, known N. Applies to any neighboring type.
                    (_, 1, Some(_)) => wrap!(0),

                    // one category, unknown N. The sensitivity here is really zero-- artificially raised
                    (Substitute, 1, None) => wrap!(1),
                    // two categories, known N. Knowing N determines the second category
                    (Substitute, 2, Some(_)) => wrap!(1),

                    // one category, unknown N
                    (AddRemove, 1, None) => wrap!(1),
                    // two categories, known N
                    (AddRemove, 2, Some(_)) => wrap!(1),

                    // over two categories, N either known or unknown. Record may switch from one bin to another.
                    (Substitute, _, _) => match k {
                        1 => wrap!(2),
                        2 => wrap!(2.0_f64.sqrt()),
                        _ => Err("KNorm sensitivity is only supported in L1 and L2 spaces".into())
                    } ,
                    // over two categories, N either known or unknown. Only one bin may be edited.
                    (AddRemove, _, _) => wrap!(1),
                }

            },
            _ => Err("Histogram sensitivity is only implemented for KNorm".into())
        }
    }
}