1// Copyright 2014 The Chromium Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style license that can be
3// found in the LICENSE file.
4
5#include "components/rappor/rappor_metric.h"
6
7#include "base/logging.h"
8#include "base/rand_util.h"
9
10namespace rappor {
11
12RapporMetric::RapporMetric(const std::string& metric_name,
13                           const RapporParameters& parameters,
14                           int32_t cohort_seed)
15    : metric_name_(metric_name),
16      parameters_(parameters),
17      sample_count_(0),
18      bloom_filter_(parameters.bloom_filter_size_bytes,
19                    parameters.bloom_filter_hash_function_count,
20                    (cohort_seed % parameters.num_cohorts) *
21                        parameters.bloom_filter_hash_function_count) {
22  DCHECK_GE(cohort_seed, 0);
23  DCHECK_LT(cohort_seed, RapporParameters::kMaxCohorts);
24}
25
26RapporMetric::~RapporMetric() {}
27
28void RapporMetric::AddSample(const std::string& str) {
29  ++sample_count_;
30  // Replace the previous sample with a 1 in sample_count_ chance so that each
31  // sample has equal probability of being reported.
32  if (base::RandGenerator(sample_count_) == 0) {
33    bloom_filter_.SetString(str);
34  }
35}
36
37ByteVector RapporMetric::GetReport(const std::string& secret) const {
38  // Generate a deterministically random mask of fake data using the
39  // client's secret key + real data as a seed.  The inclusion of the secret
40  // in the seed avoids correlations between real and fake data.
41  // The seed isn't a human-readable string.
42  const std::string personalization_string = metric_name_ +
43      std::string(bytes().begin(), bytes().end());
44  HmacByteVectorGenerator hmac_generator(bytes().size(), secret,
45                                         personalization_string);
46  const ByteVector fake_mask =
47      hmac_generator.GetWeightedRandomByteVector(parameters().fake_prob);
48  ByteVector fake_bits =
49      hmac_generator.GetWeightedRandomByteVector(parameters().fake_one_prob);
50
51  // Redact most of the real data by replacing it with the fake data, hiding
52  // and limiting the amount of information an individual client reports on.
53  const ByteVector* fake_and_redacted_bits =
54      ByteVectorMerge(fake_mask, bytes(), &fake_bits);
55
56  // Generate biased coin flips for each bit.
57  ByteVectorGenerator coin_generator(bytes().size());
58  const ByteVector zero_coins =
59      coin_generator.GetWeightedRandomByteVector(parameters().zero_coin_prob);
60  ByteVector one_coins =
61      coin_generator.GetWeightedRandomByteVector(parameters().one_coin_prob);
62
63  // Create a randomized response report on the fake and redacted data, sending
64  // the outcome of flipping a zero coin for the zero bits in that data, and of
65  // flipping a one coin for the one bits in that data, as the final report.
66  return *ByteVectorMerge(*fake_and_redacted_bits, zero_coins, &one_coins);
67}
68
69void RapporMetric::SetBytesForTesting(const ByteVector& bytes) {
70  bloom_filter_.SetBytesForTesting(bytes);
71}
72
73}  // namespace rappor
74