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