chrome/renderer/safe_browsing/phishing_term_feature_extractor_unittest.cc

   1 // Copyright (c) 2012 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 "chrome/renderer/safe_browsing/phishing_term_feature_extractor.h"
   6
   7 #include <string>
   8
   9 #include "base/bind.h"
  10 #include "base/callback.h"
  11 #include "base/containers/hash_tables.h"
  12 #include "base/location.h"
  13 #include "base/memory/scoped_ptr.h"
  14 #include "base/message_loop/message_loop.h"
  15 #include "base/single_thread_task_runner.h"
  16 #include "base/strings/string16.h"
  17 #include "base/strings/stringprintf.h"
  18 #include "base/strings/utf_string_conversions.h"
  19 #include "base/time/time.h"
  20 #include "chrome/renderer/safe_browsing/features.h"
  21 #include "chrome/renderer/safe_browsing/mock_feature_extractor_clock.h"
  22 #include "chrome/renderer/safe_browsing/murmurhash3_util.h"
  23 #include "chrome/renderer/safe_browsing/test_utils.h"
  24 #include "crypto/sha2.h"
  25 #include "testing/gmock/include/gmock/gmock.h"
  26 #include "testing/gtest/include/gtest/gtest.h"
  27
  28 using base::ASCIIToUTF16;
  29 using ::testing::Return;
  30
  31
  32 static const uint32 kMurmurHash3Seed = 2777808611U;
  33
  34 namespace safe_browsing {
  35
  36 class PhishingTermFeatureExtractorTest : public ::testing::Test {
  37  protected:
  38   void SetUp() override {
  39     base::hash_set<std::string> terms;
  40     terms.insert("one");
  41     terms.insert("one one");
  42     terms.insert("two");
  43     terms.insert("multi word test");
  44     terms.insert("capitalization");
  45     terms.insert("space");
  46     terms.insert("separator");
  47     terms.insert("punctuation");
  48     // Chinese (translation of "hello")
  49     terms.insert("\xe4\xbd\xa0\xe5\xa5\xbd");
  50     // Chinese (translation of "goodbye")
  51     terms.insert("\xe5\x86\x8d\xe8\xa7\x81");
  52
  53     for (base::hash_set<std::string>::iterator it = terms.begin();
  54          it != terms.end(); ++it) {
  55       term_hashes_.insert(crypto::SHA256HashString(*it));
  56     }
  57
  58     base::hash_set<std::string> words;
  59     words.insert("one");
  60     words.insert("two");
  61     words.insert("multi");
  62     words.insert("word");
  63     words.insert("test");
  64     words.insert("capitalization");
  65     words.insert("space");
  66     words.insert("separator");
  67     words.insert("punctuation");
  68     words.insert("\xe4\xbd\xa0\xe5\xa5\xbd");
  69     words.insert("\xe5\x86\x8d\xe8\xa7\x81");
  70
  71     for (base::hash_set<std::string>::iterator it = words.begin();
  72          it != words.end(); ++it) {
  73       word_hashes_.insert(MurmurHash3String(*it, kMurmurHash3Seed));
  74     }
  75
  76     ResetExtractor(3 /* max shingles per page */);
  77   }
  78
  79   void ResetExtractor(size_t max_shingles_per_page) {
  80     extractor_.reset(new PhishingTermFeatureExtractor(
  81         &term_hashes_,
  82         &word_hashes_,
  83         3 /* max_words_per_term */,
  84         kMurmurHash3Seed,
  85         max_shingles_per_page,
  86         4 /* shingle_size */,
  87         &clock_));
  88   }
  89
  90   // Runs the TermFeatureExtractor on |page_text|, waiting for the
  91   // completion callback.  Returns the success boolean from the callback.
  92   bool ExtractFeatures(const base::string16* page_text,
  93                        FeatureMap* features,
  94                        std::set<uint32>* shingle_hashes) {
  95     success_ = false;
  96     extractor_->ExtractFeatures(
  97         page_text,
  98         features,
  99         shingle_hashes,
 100         base::Bind(&PhishingTermFeatureExtractorTest::ExtractionDone,
 101                    base::Unretained(this)));
 102     msg_loop_.Run();
 103     return success_;
 104   }
 105
 106   void PartialExtractFeatures(const base::string16* page_text,
 107                               FeatureMap* features,
 108                               std::set<uint32>* shingle_hashes) {
 109     extractor_->ExtractFeatures(
 110         page_text,
 111         features,
 112         shingle_hashes,
 113         base::Bind(&PhishingTermFeatureExtractorTest::ExtractionDone,
 114                    base::Unretained(this)));
 115     msg_loop_.task_runner()->PostTask(
 116         FROM_HERE, base::Bind(&PhishingTermFeatureExtractorTest::QuitExtraction,
 117                               base::Unretained(this)));
 118     msg_loop_.RunUntilIdle();
 119   }
 120
 121   // Completion callback for feature extraction.
 122   void ExtractionDone(bool success) {
 123     success_ = success;
 124     msg_loop_.Quit();
 125   }
 126
 127   void QuitExtraction() {
 128     extractor_->CancelPendingExtraction();
 129     msg_loop_.Quit();
 130   }
 131
 132   base::MessageLoop msg_loop_;
 133   MockFeatureExtractorClock clock_;
 134   scoped_ptr<PhishingTermFeatureExtractor> extractor_;
 135   base::hash_set<std::string> term_hashes_;
 136   base::hash_set<uint32> word_hashes_;
 137   bool success_;  // holds the success value from ExtractFeatures
 138 };
 139
 140 TEST_F(PhishingTermFeatureExtractorTest, ExtractFeatures) {
 141   // This test doesn't exercise the extraction timing.
 142   EXPECT_CALL(clock_, Now()).WillRepeatedly(Return(base::TimeTicks::Now()));
 143
 144   base::string16 page_text = ASCIIToUTF16("blah");
 145   FeatureMap expected_features;  // initially empty
 146   std::set<uint32> expected_shingle_hashes;
 147
 148   FeatureMap features;
 149   std::set<uint32> shingle_hashes;
 150   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 151   ExpectFeatureMapsAreEqual(features, expected_features);
 152   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 153
 154   page_text = ASCIIToUTF16("one one");
 155   expected_features.Clear();
 156   expected_features.AddBooleanFeature(features::kPageTerm +
 157                                       std::string("one"));
 158   expected_features.AddBooleanFeature(features::kPageTerm +
 159                                       std::string("one one"));
 160   expected_shingle_hashes.clear();
 161
 162   features.Clear();
 163   shingle_hashes.clear();
 164   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 165   ExpectFeatureMapsAreEqual(features, expected_features);
 166   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 167
 168   page_text = ASCIIToUTF16("bla bla multi word test bla");
 169   expected_features.Clear();
 170   expected_features.AddBooleanFeature(features::kPageTerm +
 171                                       std::string("multi word test"));
 172   expected_shingle_hashes.clear();
 173   expected_shingle_hashes.insert(MurmurHash3String("bla bla multi word ",
 174                                                    kMurmurHash3Seed));
 175   expected_shingle_hashes.insert(MurmurHash3String("bla multi word test ",
 176                                                    kMurmurHash3Seed));
 177   expected_shingle_hashes.insert(MurmurHash3String("multi word test bla ",
 178                                                    kMurmurHash3Seed));
 179
 180   features.Clear();
 181   shingle_hashes.clear();
 182   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 183   ExpectFeatureMapsAreEqual(features, expected_features);
 184   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 185
 186   // This text has all of the words for one of the terms, but they are
 187   // not in the correct order.
 188   page_text = ASCIIToUTF16("bla bla test word multi bla");
 189   expected_features.Clear();
 190   expected_shingle_hashes.clear();
 191   expected_shingle_hashes.insert(MurmurHash3String("bla bla test word ",
 192                                                    kMurmurHash3Seed));
 193   expected_shingle_hashes.insert(MurmurHash3String("bla test word multi ",
 194                                                    kMurmurHash3Seed));
 195   expected_shingle_hashes.insert(MurmurHash3String("test word multi bla ",
 196                                                    kMurmurHash3Seed));
 197
 198   features.Clear();
 199   shingle_hashes.clear();
 200   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 201   ExpectFeatureMapsAreEqual(features, expected_features);
 202   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 203
 204   // Test various separators.
 205   page_text = ASCIIToUTF16("Capitalization plus non-space\n"
 206                            "separator... punctuation!");
 207   expected_features.Clear();
 208   expected_features.AddBooleanFeature(features::kPageTerm +
 209                                       std::string("capitalization"));
 210   expected_features.AddBooleanFeature(features::kPageTerm +
 211                                       std::string("space"));
 212   expected_features.AddBooleanFeature(features::kPageTerm +
 213                                       std::string("separator"));
 214   expected_features.AddBooleanFeature(features::kPageTerm +
 215                                       std::string("punctuation"));
 216   expected_shingle_hashes.clear();
 217   expected_shingle_hashes.insert(
 218       MurmurHash3String("capitalization plus non space ", kMurmurHash3Seed));
 219   expected_shingle_hashes.insert(MurmurHash3String("plus non space separator ",
 220                                                    kMurmurHash3Seed));
 221   expected_shingle_hashes.insert(
 222       MurmurHash3String("non space separator punctuation ", kMurmurHash3Seed));
 223
 224   features.Clear();
 225   shingle_hashes.clear();
 226   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 227   ExpectFeatureMapsAreEqual(features, expected_features);
 228   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 229
 230   // Test a page with too many words and we should only 3 minimum hashes.
 231   page_text = ASCIIToUTF16("This page has way too many words.");
 232   expected_features.Clear();
 233   expected_shingle_hashes.clear();
 234   expected_shingle_hashes.insert(MurmurHash3String("this page has way ",
 235                                                    kMurmurHash3Seed));
 236   expected_shingle_hashes.insert(MurmurHash3String("page has way too ",
 237                                                    kMurmurHash3Seed));
 238   expected_shingle_hashes.insert(MurmurHash3String("has way too many ",
 239                                                    kMurmurHash3Seed));
 240   expected_shingle_hashes.insert(MurmurHash3String("way too many words ",
 241                                                    kMurmurHash3Seed));
 242   std::set<uint32>::iterator it = expected_shingle_hashes.end();
 243   expected_shingle_hashes.erase(--it);
 244
 245   features.Clear();
 246   shingle_hashes.clear();
 247   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 248   ExpectFeatureMapsAreEqual(features, expected_features);
 249   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 250
 251   // Test with empty page text.
 252   page_text = base::string16();
 253   expected_features.Clear();
 254   expected_shingle_hashes.clear();
 255   features.Clear();
 256   shingle_hashes.clear();
 257   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 258   ExpectFeatureMapsAreEqual(features, expected_features);
 259   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 260
 261 #if !defined(OS_ANDROID)
 262   // The test code is disabled due to http://crbug.com/392234
 263   // The client-side detection feature is not enabled on Android yet.
 264   // If we decided to enable the feature, we need to fix the bug first.
 265
 266   // Chinese translation of the phrase "hello goodbye hello goodbye". This tests
 267   // that we can correctly separate terms in languages that don't use spaces.
 268   page_text =
 269       base::UTF8ToUTF16("\xe4\xbd\xa0\xe5\xa5\xbd\xe5\x86\x8d\xe8\xa7\x81"
 270                         "\xe4\xbd\xa0\xe5\xa5\xbd\xe5\x86\x8d\xe8\xa7\x81");
 271   expected_features.Clear();
 272   expected_features.AddBooleanFeature(
 273       features::kPageTerm + std::string("\xe4\xbd\xa0\xe5\xa5\xbd"));
 274   expected_features.AddBooleanFeature(
 275       features::kPageTerm + std::string("\xe5\x86\x8d\xe8\xa7\x81"));
 276   expected_shingle_hashes.clear();
 277   expected_shingle_hashes.insert(MurmurHash3String(
 278       "\xe4\xbd\xa0\xe5\xa5\xbd \xe5\x86\x8d\xe8\xa7\x81 "
 279       "\xe4\xbd\xa0\xe5\xa5\xbd \xe5\x86\x8d\xe8\xa7\x81 ", kMurmurHash3Seed));
 280
 281   features.Clear();
 282   shingle_hashes.clear();
 283   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 284   ExpectFeatureMapsAreEqual(features, expected_features);
 285   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 286 #endif
 287 }
 288
 289 TEST_F(PhishingTermFeatureExtractorTest, Continuation) {
 290   // For this test, we'll cause the feature extraction to run multiple
 291   // iterations by incrementing the clock.
 292   ResetExtractor(200 /* max shingles per page */);
 293
 294   // This page has a total of 30 words.  For the features to be computed
 295   // correctly, the extractor has to process the entire string of text.
 296   base::string16 page_text(ASCIIToUTF16("one "));
 297   for (int i = 0; i < 28; ++i) {
 298     page_text.append(ASCIIToUTF16(base::StringPrintf("%d ", i)));
 299   }
 300   page_text.append(ASCIIToUTF16("two"));
 301
 302   // Advance the clock 3 ms every 5 words processed, 10 ms between chunks.
 303   // Note that this assumes kClockCheckGranularity = 5 and
 304   // kMaxTimePerChunkMs = 10.
 305   base::TimeTicks now = base::TimeTicks::Now();
 306   EXPECT_CALL(clock_, Now())
 307       // Time check at the start of extraction.
 308       .WillOnce(Return(now))
 309       // Time check at the start of the first chunk of work.
 310       .WillOnce(Return(now))
 311       // Time check after the first 5 words.
 312       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(3)))
 313       // Time check after the next 5 words.
 314       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(6)))
 315       // Time check after the next 5 words.
 316       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(9)))
 317       // Time check after the next 5 words.  This is over the chunk
 318       // time limit, so a continuation task will be posted.
 319       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(12)))
 320       // Time check at the start of the second chunk of work.
 321       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(22)))
 322       // Time check after the next 5 words.
 323       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(25)))
 324       // Time check after the next 5 words.
 325       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(28)))
 326       // A final check for the histograms.
 327       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(30)));
 328
 329   FeatureMap expected_features;
 330   expected_features.AddBooleanFeature(features::kPageTerm +
 331                                       std::string("one"));
 332   expected_features.AddBooleanFeature(features::kPageTerm +
 333                                       std::string("two"));
 334   std::set<uint32> expected_shingle_hashes;
 335   expected_shingle_hashes.insert(
 336       MurmurHash3String("one 0 1 2 ", kMurmurHash3Seed));
 337   expected_shingle_hashes.insert(
 338       MurmurHash3String("0 1 2 3 ", kMurmurHash3Seed));
 339   expected_shingle_hashes.insert(
 340       MurmurHash3String("1 2 3 4 ", kMurmurHash3Seed));
 341   expected_shingle_hashes.insert(
 342       MurmurHash3String("2 3 4 5 ", kMurmurHash3Seed));
 343   expected_shingle_hashes.insert(
 344       MurmurHash3String("3 4 5 6 ", kMurmurHash3Seed));
 345   expected_shingle_hashes.insert(
 346       MurmurHash3String("4 5 6 7 ", kMurmurHash3Seed));
 347   expected_shingle_hashes.insert(
 348       MurmurHash3String("5 6 7 8 ", kMurmurHash3Seed));
 349   expected_shingle_hashes.insert(
 350       MurmurHash3String("6 7 8 9 ", kMurmurHash3Seed));
 351   expected_shingle_hashes.insert(
 352       MurmurHash3String("7 8 9 10 ", kMurmurHash3Seed));
 353   expected_shingle_hashes.insert(
 354       MurmurHash3String("8 9 10 11 ", kMurmurHash3Seed));
 355   expected_shingle_hashes.insert(
 356       MurmurHash3String("9 10 11 12 ", kMurmurHash3Seed));
 357   expected_shingle_hashes.insert(
 358       MurmurHash3String("10 11 12 13 ", kMurmurHash3Seed));
 359   expected_shingle_hashes.insert(
 360       MurmurHash3String("11 12 13 14 ", kMurmurHash3Seed));
 361   expected_shingle_hashes.insert(
 362       MurmurHash3String("12 13 14 15 ", kMurmurHash3Seed));
 363   expected_shingle_hashes.insert(
 364       MurmurHash3String("13 14 15 16 ", kMurmurHash3Seed));
 365   expected_shingle_hashes.insert(
 366       MurmurHash3String("14 15 16 17 ", kMurmurHash3Seed));
 367   expected_shingle_hashes.insert(
 368       MurmurHash3String("15 16 17 18 ", kMurmurHash3Seed));
 369   expected_shingle_hashes.insert(
 370       MurmurHash3String("16 17 18 19 ", kMurmurHash3Seed));
 371   expected_shingle_hashes.insert(
 372       MurmurHash3String("17 18 19 20 ", kMurmurHash3Seed));
 373   expected_shingle_hashes.insert(
 374       MurmurHash3String("18 19 20 21 ", kMurmurHash3Seed));
 375   expected_shingle_hashes.insert(
 376       MurmurHash3String("19 20 21 22 ", kMurmurHash3Seed));
 377   expected_shingle_hashes.insert(
 378       MurmurHash3String("20 21 22 23 ", kMurmurHash3Seed));
 379   expected_shingle_hashes.insert(
 380       MurmurHash3String("21 22 23 24 ", kMurmurHash3Seed));
 381   expected_shingle_hashes.insert(
 382       MurmurHash3String("22 23 24 25 ", kMurmurHash3Seed));
 383   expected_shingle_hashes.insert(
 384       MurmurHash3String("23 24 25 26 ", kMurmurHash3Seed));
 385   expected_shingle_hashes.insert(
 386       MurmurHash3String("24 25 26 27 ", kMurmurHash3Seed));
 387   expected_shingle_hashes.insert(
 388       MurmurHash3String("25 26 27 two ", kMurmurHash3Seed));
 389
 390   FeatureMap features;
 391   std::set<uint32> shingle_hashes;
 392   ASSERT_TRUE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 393   ExpectFeatureMapsAreEqual(features, expected_features);
 394   EXPECT_THAT(expected_shingle_hashes, testing::ContainerEq(shingle_hashes));
 395   // Make sure none of the mock expectations carry over to the next test.
 396   ::testing::Mock::VerifyAndClearExpectations(&clock_);
 397
 398   // Now repeat the test with the same text, but advance the clock faster so
 399   // that the extraction time exceeds the maximum total time for the feature
 400   // extractor.  Extraction should fail.  Note that this assumes
 401   // kMaxTotalTimeMs = 500.
 402   EXPECT_CALL(clock_, Now())
 403       // Time check at the start of extraction.
 404       .WillOnce(Return(now))
 405       // Time check at the start of the first chunk of work.
 406       .WillOnce(Return(now))
 407       // Time check after the first 5 words,
 408       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(300)))
 409       // Time check at the start of the second chunk of work.
 410       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(350)))
 411       // Time check after the next 5 words.  This is over the limit.
 412       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(600)))
 413       // A final time check for the histograms.
 414       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(620)));
 415
 416   features.Clear();
 417   shingle_hashes.clear();
 418   EXPECT_FALSE(ExtractFeatures(&page_text, &features, &shingle_hashes));
 419 }
 420
 421 TEST_F(PhishingTermFeatureExtractorTest, PartialExtractionTest) {
 422   scoped_ptr<base::string16> page_text(
 423       new base::string16(ASCIIToUTF16("one ")));
 424   for (int i = 0; i < 28; ++i) {
 425     page_text->append(ASCIIToUTF16(base::StringPrintf("%d ", i)));
 426   }
 427
 428   base::TimeTicks now = base::TimeTicks::Now();
 429   EXPECT_CALL(clock_, Now())
 430       // Time check at the start of extraction.
 431       .WillOnce(Return(now))
 432       // Time check at the start of the first chunk of work.
 433       .WillOnce(Return(now))
 434       // Time check after the first 5 words.
 435       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(7)))
 436       // Time check after the next 5 words. This should be greater than
 437       // kMaxTimePerChunkMs so that we stop and schedule extraction for later.
 438       .WillOnce(Return(now + base::TimeDelta::FromMilliseconds(14)));
 439
 440   FeatureMap features;
 441   std::set<uint32> shingle_hashes;
 442   // Extract first 10 words then stop.
 443   PartialExtractFeatures(page_text.get(), &features, &shingle_hashes);
 444
 445   page_text.reset(new base::string16());
 446   for (int i = 30; i < 58; ++i) {
 447     page_text->append(ASCIIToUTF16(base::StringPrintf("%d ", i)));
 448   }
 449   page_text->append(ASCIIToUTF16("multi word test "));
 450   features.Clear();
 451   shingle_hashes.clear();
 452
 453   // This part doesn't exercise the extraction timing.
 454   EXPECT_CALL(clock_, Now()).WillRepeatedly(Return(base::TimeTicks::Now()));
 455
 456   // Now extract normally and make sure nothing breaks.
 457   EXPECT_TRUE(ExtractFeatures(page_text.get(), &features, &shingle_hashes));
 458
 459   FeatureMap expected_features;
 460   expected_features.AddBooleanFeature(features::kPageTerm +
 461                                       std::string("multi word test"));
 462   ExpectFeatureMapsAreEqual(features, expected_features);
 463 }
 464
 465 }  // namespace safe_browsing