Extracting page shingle hashes for similarity detection.

chrome/renderer/safe_browsing/phishing_classifier.cc

Index: chrome/renderer/safe_browsing/phishing_classifier.cc
diff --git a/chrome/renderer/safe_browsing/phishing_classifier.cc b/chrome/renderer/safe_browsing/phishing_classifier.cc
index f48e6c138c74e444e79c58d8edb32809d2b6daa7..cc7e623673183c90659383b735f82d6588e9ef3b 100644
--- a/chrome/renderer/safe_browsing/phishing_classifier.cc
+++ b/chrome/renderer/safe_browsing/phishing_classifier.cc
@@ -61,8 +61,10 @@ void PhishingClassifier::set_phishing_scorer(const Scorer* scorer) {
     term_extractor_.reset(new PhishingTermFeatureExtractor(
         &scorer_->page_terms(),
         &scorer_->page_words(),
+        scorer_->max_hashes_per_page(),
         scorer_->max_words_per_term(),
         scorer_->murmurhash3_seed(),
+        scorer_->shingle_size(),
         clock_.get()));
   } else {
     // We're disabling client-side phishing detection, so tear down all
@@ -154,12 +156,14 @@ void PhishingClassifier::CancelPendingClassification() {
 }
 
 void PhishingClassifier::DOMExtractionFinished(bool success) {
+  shingle_hashes_.reset(new std::set<uint32>);
   if (success) {
     // Term feature extraction can take awhile, so it runs asynchronously
     // in several chunks of work and invokes the callback when finished.
     term_extractor_->ExtractFeatures(
         page_text_,
         features_.get(),
+        shingle_hashes_.get(),
         base::Bind(&PhishingClassifier::TermExtractionFinished,
                    base::Unretained(this)));
   } else {
@@ -197,6 +201,10 @@ void PhishingClassifier::TermExtractionFinished(bool success) {
       feature->set_name(it->first);
       feature->set_value(it->second);
     }
+    for (std::set<uint32>::const_iterator it = shingle_hashes_->begin();
+         it != shingle_hashes_->end(); ++it) {
+      verdict.add_shingle_hashes(*it);
+    }
     float score = static_cast<float>(scorer_->ComputeScore(hashed_features));
     verdict.set_client_score(score);
     verdict.set_is_phishing(score >= kPhishyThreshold);
@@ -236,6 +244,7 @@ void PhishingClassifier::Clear() {
   page_text_ = NULL;
   done_callback_.Reset();
   features_.reset(NULL);
+  shingle_hashes_.reset(NULL);
 }
 
 }  // namespace safe_browsing