tor-browser

The Tor Browser
git clone https://git.dasho.dev/tor-browser.git
Log | Files | Refs | README | LICENSE

entropy_pool_test.cc (4017B)


      1 // Copyright 2017 The Abseil Authors.
      2 //
      3 // Licensed under the Apache License, Version 2.0 (the "License");
      4 // you may not use this file except in compliance with the License.
      5 // You may obtain a copy of the License at
      6 //
      7 //      https://www.apache.org/licenses/LICENSE-2.0
      8 //
      9 // Unless required by applicable law or agreed to in writing, software
     10 // distributed under the License is distributed on an "AS IS" BASIS,
     11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
     12 // See the License for the specific language governing permissions and
     13 // limitations under the License.
     14 
     15 #include "absl/random/internal/entropy_pool.h"
     16 
     17 #include <bitset>
     18 #include <cmath>
     19 #include <cstddef>
     20 #include <cstdint>
     21 #include <thread>  // NOLINT
     22 #include <utility>
     23 #include <vector>
     24 
     25 #include "gtest/gtest.h"
     26 #include "absl/container/flat_hash_set.h"
     27 #include "absl/synchronization/mutex.h"
     28 
     29 namespace {
     30 
     31 using ::absl::random_internal::GetEntropyFromRandenPool;
     32 
     33 TEST(EntropyPoolTest, DistinctSequencesPerThread) {
     34  using result_type = uint32_t;
     35  constexpr int kNumThreads = 12;
     36  constexpr size_t kValuesPerThread = 32;
     37 
     38  // Acquire entropy from multiple threads.
     39  std::vector<std::vector<result_type>> data;
     40  {
     41    absl::Mutex mu;
     42    std::vector<std::thread> threads;
     43    for (int i = 0; i < kNumThreads; i++) {
     44      threads.emplace_back([&]() {
     45        std::vector<result_type> v(kValuesPerThread);
     46        GetEntropyFromRandenPool(v.data(), sizeof(result_type) * v.size());
     47        absl::MutexLock l(&mu);
     48        data.push_back(std::move(v));
     49      });
     50    }
     51    for (auto& t : threads) t.join();
     52  }
     53 
     54  EXPECT_EQ(data.size(), kNumThreads);
     55 
     56  // There should be essentially no duplicates in the sequences.
     57  size_t expected_size = 0;
     58  absl::flat_hash_set<result_type> seen;
     59  for (const auto& v : data) {
     60    expected_size += v.size();
     61    for (result_type x : v) seen.insert(x);
     62  }
     63  EXPECT_GE(seen.size(), expected_size - 1);
     64 }
     65 
     66 // This validates that sequences are independent.
     67 TEST(EntropyPoolTest, ValidateDistribution) {
     68  using result_type = uint32_t;
     69  constexpr int kNumOutputs = 16;
     70  std::vector<result_type> a(kNumOutputs);
     71  std::vector<result_type> b(kNumOutputs);
     72 
     73  GetEntropyFromRandenPool(a.data(), sizeof(a[0]) * a.size());
     74  GetEntropyFromRandenPool(b.data(), sizeof(b[0]) * b.size());
     75 
     76  // Compare the two sequences, counting the number of bits that are different,
     77  // then verify using a normal-approximation of the binomial distribution.
     78  size_t changed_bits = 0;
     79  size_t total_set = 0;
     80  size_t equal_count = 0;
     81  size_t zero_count = 0;
     82  for (size_t i = 0; i < a.size(); ++i) {
     83    std::bitset<sizeof(result_type) * 8> changed_set(a[i] ^ b[i]);
     84    changed_bits += changed_set.count();
     85 
     86    std::bitset<sizeof(result_type) * 8> a_set(a[i]);
     87    std::bitset<sizeof(result_type) * 8> b_set(b[i]);
     88    total_set += a_set.count() + b_set.count();
     89 
     90    equal_count += (a[i] == b[i]) ? 1 : 0;
     91 
     92    zero_count += (a[i] == 0) ? 1 : 0;
     93    zero_count += (b[i] == 0) ? 1 : 0;
     94  }
     95 
     96  constexpr size_t kNBits = kNumOutputs * sizeof(result_type) * 8;
     97 
     98  // This should be a binomial distribution with:
     99  //    p = 0.5
    100  //    n = kNBits
    101  //    sigma =~ 11.3 (sqrt(n * 0.5 * 0.5))
    102  // So we expect the number of changed bits to be within 5 standard deviations
    103  // of the mean; this should fail less than one in 3 million times.
    104  EXPECT_NEAR(changed_bits, kNBits * 0.5, 5 * std::sqrt(kNBits))
    105      << "@" << changed_bits / static_cast<double>(kNBits);
    106 
    107  // Verify that the number of set bits is also within the expected range;
    108  // Note that this is summed over the two sequences, so the number of trials
    109  // is twice the number of bits.
    110  EXPECT_NEAR(total_set, kNBits, 5 * std::sqrt(2 * kNBits))
    111      << "@" << total_set / static_cast<double>(2 * kNBits);
    112 
    113  // A[i] == B[i] with probability ~= 16 * 1/2^32; certainly less than 1.
    114  EXPECT_LE(equal_count, 1);
    115 
    116  // Zeros values must be rare; 32 / 2^32 is certainly less than 1.
    117  EXPECT_LE(zero_count, 1);
    118 }
    119 }  // namespace