Refactor the captive portal code to move from the ssl_blocking_page class to the ssl_error_classific

chrome/browser/ssl/ssl_error_classification.cc

Index: chrome/browser/ssl/ssl_error_classification.cc
diff --git a/chrome/browser/ssl/ssl_error_classification.cc b/chrome/browser/ssl/ssl_error_classification.cc
index 0f8911dac843bcaaf7ea553afdbadfc743ac6b52..e6464d8c8bace76a0c196e39c06bd1d6e1e2d979 100644
--- a/chrome/browser/ssl/ssl_error_classification.cc
+++ b/chrome/browser/ssl/ssl_error_classification.cc
@@ -12,21 +12,31 @@
 #include "base/strings/string_split.h"
 #include "base/strings/utf_string_conversions.h"
 #include "base/time/time.h"
+#include "chrome/browser/browser_process.h"
+#include "chrome/browser/chrome_notification_types.h"
+#include "chrome/browser/profiles/profile.h"
 #include "chrome/browser/ssl/ssl_error_info.h"
+#include "content/public/browser/notification_service.h"
+#include "content/public/browser/web_contents.h"
 #include "net/base/net_util.h"
 #include "net/base/registry_controlled_domains/registry_controlled_domain.h"
 #include "net/cert/x509_cert_types.h"
 #include "net/cert/x509_certificate.h"
 #include "url/gurl.h"
 
-using base::Time;
-using base::TimeTicks;
-using base::TimeDelta;
+#if defined(ENABLE_CAPTIVE_PORTAL_DETECTION)
+#include "chrome/browser/captive_portal/captive_portal_service.h"
+#include "chrome/browser/captive_portal/captive_portal_service_factory.h"
+#endif
 
 #if defined(OS_WIN)
 #include "base/win/windows_version.h"
 #endif
 
+using base::Time;
+using base::TimeTicks;
+using base::TimeDelta;
+
 namespace {
 
 // Events for UMA. Do not reorder or change!
@@ -38,9 +48,23 @@ enum SSLInterstitialCause {
   SUBDOMAIN_INVERSE_MATCH,
   SUBDOMAIN_OUTSIDE_WILDCARD,
   HOST_NAME_NOT_KNOWN_TLD,
+  LIKELY_MULTI_TENANT_HOSTING,
   UNUSED_INTERSTITIAL_CAUSE_ENTRY,
 };
 
+// Events for UMA. Do not reorder or change!
+enum SSLInterstitialCauseCaptivePortal {
+  CAPTIVE_PORTAL_DETECTION_ENABLED,
+  CAPTIVE_PORTAL_DETECTION_ENABLED_OVERRIDABLE,
+  CAPTIVE_PORTAL_PROBE_COMPLETED,
+  CAPTIVE_PORTAL_PROBE_COMPLETED_OVERRIDABLE,
+  CAPTIVE_PORTAL_NO_RESPONSE,
+  CAPTIVE_PORTAL_NO_RESPONSE_OVERRIDABLE,
+  CAPTIVE_PORTAL_DETECTED,
+  CAPTIVE_PORTAL_DETECTED_OVERRIDABLE,
+  UNUSED_CAPTIVE_PORTAL_EVENT,
+};
+
 // Scores/weights which will be constant through all the SSL error types.
 static const float kServerWeight = 0.5f;
 static const float kClientWeight = 0.5f;
@@ -55,26 +79,104 @@ void RecordSSLInterstitialCause(bool overridable, SSLInterstitialCause event) {
   }
 }
 
+void RecordCaptivePortalEventStats(SSLInterstitialCauseCaptivePortal event) {
+  UMA_HISTOGRAM_ENUMERATION("interstitial.ssl.captive_portal",
+                            event,
+                            UNUSED_CAPTIVE_PORTAL_EVENT);
+}
+
+int GetLevensteinDistance(const std::string& str1,
+                          const std::string& str2) {
+  if (str1 == str2)
+    return 0;
+  if (str1.size() == 0)
+    return str2.size();
+  if (str2.size() == 0)
+    return str1.size();
+  std::vector<int> kFirstRow(str2.size() + 1, 0);
+  std::vector<int> kSecondRow(str2.size() + 1, 0);
+
+  for (size_t i = 0; i < kFirstRow.size(); ++i)
+    kFirstRow[i] = i;
+  for (size_t i = 0; i < str1.size(); ++i) {
+    kSecondRow[0] = i + 1;
+    for (size_t j = 0; j < str2.size(); ++j) {
+      int cost = str1[i] == str2[j] ? 0 : 1;
+      kSecondRow[j+1] = std::min(std::min(
+          kSecondRow[j] + 1, kFirstRow[j + 1] + 1), kFirstRow[j] + cost);
+    }
+    for (size_t j = 0; j < kFirstRow.size(); j++)
+      kFirstRow[j] = kSecondRow[j];
+  }
+  return kSecondRow[str2.size()];
+}
+
 } // namespace
 
 SSLErrorClassification::SSLErrorClassification(
+    content::WebContents* web_contents,
     const base::Time& current_time,
     const GURL& url,
+    int cert_error,
     const net::X509Certificate& cert)
-  : current_time_(current_time),
+  : web_contents_(web_contents),
+    current_time_(current_time),
     request_url_(url),
-    cert_(cert) { }
+    cert_error_(cert_error),
+    cert_(cert),
+    captive_portal_detection_enabled_(false),
+    captive_portal_probe_completed_(false),
+    captive_portal_no_response_(false),
+    captive_portal_detected_(false) {
+#if defined(ENABLE_CAPTIVE_PORTAL_DETECTION)
+  Profile* profile = Profile::FromBrowserContext(
+      web_contents_->GetBrowserContext());
+  CaptivePortalService* captive_portal_service =
+      CaptivePortalServiceFactory::GetForProfile(profile);
+  captive_portal_detection_enabled_ = captive_portal_service->enabled();
+  captive_portal_service->DetectCaptivePortal();
+  registrar_.Add(this,
+                 chrome::NOTIFICATION_CAPTIVE_PORTAL_CHECK_RESULT,
+                 content::Source<Profile>(profile));
+#endif
+}
 
 SSLErrorClassification::~SSLErrorClassification() { }
 
-float SSLErrorClassification::InvalidDateSeverityScore(
-    int cert_error) const {
+void SSLErrorClassification::RecordCaptivePortalUMAStatistics(
+    bool overridable) const {
+#if defined(ENABLE_CAPTIVE_PORTAL_DETECTION)
+  if (captive_portal_detection_enabled_)
+    RecordCaptivePortalEventStats(
+        overridable ?
+        CAPTIVE_PORTAL_DETECTION_ENABLED_OVERRIDABLE :
+        CAPTIVE_PORTAL_DETECTION_ENABLED);
+  if (captive_portal_probe_completed_)
+    RecordCaptivePortalEventStats(
+        overridable ?
+        CAPTIVE_PORTAL_PROBE_COMPLETED_OVERRIDABLE :
+        CAPTIVE_PORTAL_PROBE_COMPLETED);
+  // Log only one of portal detected and no response results.
+  if (captive_portal_detected_)
+    RecordCaptivePortalEventStats(
+        overridable ?
+        CAPTIVE_PORTAL_DETECTED_OVERRIDABLE :
+        CAPTIVE_PORTAL_DETECTED);
+  else if (captive_portal_no_response_)
+    RecordCaptivePortalEventStats(
+        overridable ?
+        CAPTIVE_PORTAL_NO_RESPONSE_OVERRIDABLE :
+        CAPTIVE_PORTAL_NO_RESPONSE);
+#endif
+}
+
+void SSLErrorClassification::InvalidDateSeverityScore() {
   SSLErrorInfo::ErrorType type =
-      SSLErrorInfo::NetErrorToErrorType(cert_error);
+      SSLErrorInfo::NetErrorToErrorType(cert_error_);
   DCHECK(type == SSLErrorInfo::CERT_DATE_INVALID);
+
   // Client-side characteristics. Check whether or not the system's clock is
-  // wrong and whether or not the user has already encountered this error
-  // before.
+  // wrong and whether or not the user has encountered this error before.
   float severity_date_score = 0.0f;
 
   static const float kCertificateExpiredWeight = 0.3f;
@@ -103,19 +205,20 @@ float SSLErrorClassification::InvalidDateSeverityScore(
   }
   if (current_time_ < cert_.valid_start())
     severity_date_score += kServerWeight * kNotYetValidWeight;
-  return severity_date_score;
+  // TODO(radhikabhar): Record the severity score in a histogram. This will be
+  // in the next CL - just called the function in ssl_blocking_page.cc.
 }
 
-float SSLErrorClassification::InvalidCommonNameSeverityScore(
-    int cert_error) const {
+void SSLErrorClassification::InvalidCommonNameSeverityScore() {
   SSLErrorInfo::ErrorType type =
-      SSLErrorInfo::NetErrorToErrorType(cert_error);
+      SSLErrorInfo::NetErrorToErrorType(cert_error_);
   DCHECK(type == SSLErrorInfo::CERT_COMMON_NAME_INVALID);
   float severity_name_score = 0.0f;
 
   static const float kWWWDifferenceWeight = 0.3f;
-  static const float kSubDomainWeight = 0.2f;
-  static const float kSubDomainInverseWeight = 1.0f;
+  static const float kNameUnderAnyNamesWeight = 0.2f;
+  static const float kAnyNamesUnderNamesWeight = 1.0f;
+  static const float kLikelyMultiTenantHostingWeight = 0.1f;
 
   std::string host_name = request_url_.host();
   if (IsHostNameKnownTLD(host_name)) {
@@ -129,18 +232,27 @@ float SSLErrorClassification::InvalidCommonNameSeverityScore(
     cert_.GetDNSNames(&dns_names);
     std::vector<Tokens> dns_name_tokens = GetTokenizedDNSNames(dns_names);
     if (NameUnderAnyNames(host_name_tokens, dns_name_tokens))
-      severity_name_score += kServerWeight * kSubDomainWeight;
+      severity_name_score += kServerWeight * kNameUnderAnyNamesWeight;
     // Inverse case is more likely to be a MITM attack.
     if (AnyNamesUnderName(dns_name_tokens, host_name_tokens))
-      severity_name_score += kServerWeight * kSubDomainInverseWeight;
+      severity_name_score += kServerWeight * kAnyNamesUnderNamesWeight;
+    if (IsCertLikelyFromMultiTenantHosting())
+      severity_name_score += kServerWeight * kLikelyMultiTenantHostingWeight;
   }
-  return severity_name_score;
+
+  static const float kEnvironmentWeight = 0.25f;
+
+  severity_name_score += kClientWeight * kEnvironmentWeight *
+      CalculateScoreEnvironments();
+  // TODO(radhikabhar): Record the severity score in a histogram. Same as above
+  // - this will be in the next CL. So just called the function in the
+  // ssl_blocking_page.cc.
 }
 
-void SSLErrorClassification::RecordUMAStatistics(bool overridable,
-                                                 int cert_error) {
+void SSLErrorClassification::RecordUMAStatistics(
+    bool overridable) const {
   SSLErrorInfo::ErrorType type =
-      SSLErrorInfo::NetErrorToErrorType(cert_error);
+      SSLErrorInfo::NetErrorToErrorType(cert_error_);
   switch (type) {
     case SSLErrorInfo::CERT_DATE_INVALID: {
       if (IsUserClockInThePast(base::Time::NowFromSystemTime()))
@@ -164,6 +276,8 @@ void SSLErrorClassification::RecordUMAStatistics(bool overridable,
           RecordSSLInterstitialCause(overridable, SUBDOMAIN_MATCH);
         if (AnyNamesUnderName(dns_name_tokens, host_name_tokens))
           RecordSSLInterstitialCause(overridable, SUBDOMAIN_INVERSE_MATCH);
+        if (IsCertLikelyFromMultiTenantHosting())
+          RecordSSLInterstitialCause(overridable, LIKELY_MULTI_TENANT_HOSTING);
       } else {
          RecordSSLInterstitialCause(overridable, HOST_NAME_NOT_KNOWN_TLD);
       }
@@ -196,6 +310,32 @@ float SSLErrorClassification::CalculateScoreTimePassedSinceExpiry() const {
     return kLowThresholdWeight;
 }
 
+float SSLErrorClassification::CalculateScoreEnvironments() const {
+  static const float kWifiWeight = 0.7f;
+  static const float kCellularWeight = 0.7f;
+  static const float kHotspotWeight = 0.2f;
+  static const float kEthernetWeight = 0.7f;
+  static const float kOtherWeight = 0.7f;
+  net::NetworkChangeNotifier::ConnectionType type =
+      net::NetworkChangeNotifier::GetConnectionType();
+  if (type == net::NetworkChangeNotifier::CONNECTION_WIFI)
+    return kWifiWeight;
+  if (type == net::NetworkChangeNotifier::CONNECTION_2G ||
+      type == net::NetworkChangeNotifier::CONNECTION_3G ||
+      type == net::NetworkChangeNotifier::CONNECTION_4G ) {
+    return kCellularWeight;
+  }
+  if (type == net::NetworkChangeNotifier::CONNECTION_ETHERNET)
+    return kEthernetWeight;
+#if defined(ENABLE_CAPTIVE_PORTAL_DETECTION)
+  // Assume if captive portals are detected then the user is connected using a
+  // hot spot.
+  if (captive_portal_probe_completed_ && captive_portal_detected_)
+      return kHotspotWeight;
+#endif
+  return kOtherWeight;
+}
+
 bool SSLErrorClassification::IsUserClockInThePast(const base::Time& time_now) {
   base::Time build_time = base::GetBuildTime();
   if (time_now < build_time - base::TimeDelta::FromDays(2))
@@ -366,3 +506,84 @@ bool SSLErrorClassification::IsSubDomainOutsideWildcard(
   }
   return result;
 }
+
+bool SSLErrorClassification::IsCertLikelyFromMultiTenantHosting() const {
+  std::string host_name = request_url_.host();
+  std::vector<std::string> dns_names;
+  std::vector<std::string> dns_names_domain;
+  cert_.GetDNSNames(&dns_names);
+  size_t dns_names_size = dns_names.size();
+
+  // If there is only 1 DNS name then it is definitely not a shared certificate.
+  if (dns_names_size == 0 || dns_names_size == 1)
+    return false;
+
+  // Check to see if all the domains in the SAN field in the SSL certificate are
+  // the same or not.
+  for (size_t i = 0; i < dns_names_size; ++i) {
+    dns_names_domain.push_back(
+        net::registry_controlled_domains::
+        GetDomainAndRegistry(
+            dns_names[i],
+            net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES));
+  }
+  for (size_t i = 1; i < dns_names_domain.size(); ++i) {
+    if (dns_names_domain[i] != dns_names_domain[0])
+      return false;
+  }
+
+  // If the number of DNS names is more than 5 then assume that it is a shared
+  // certificate.
+  static const int kDistinctNameThreshold = 5;
+  if (dns_names_size > kDistinctNameThreshold)
+    return true;
+
+  // Heuristic - The edit distance between all the strings should be at least 5
+  // for it to be counted as a shared SSLCertificate. If even one pair of
+  // strings edit distance is below 5 then the certificate is no longer
+  // considered as a shared certificate. Include the host name in the URL also
+  // while comparing.
+  dns_names.push_back(host_name);
+  static const int kMinimumEditDsitance = 5;
+  for (size_t i = 0; i < dns_names_size; ++i) {
+    for (size_t j = i + 1; j < dns_names_size; ++j) {
+      int edit_distance = GetLevensteinDistance(dns_names[i], dns_names[j]);
+      if (edit_distance < kMinimumEditDsitance)
+        return false;
+    }
+  }
+  return true;
+}
+
+
+void SSLErrorClassification::Observe(
+    int type,
+    const content::NotificationSource& source,
+    const content::NotificationDetails& details) {
+#if defined(ENABLE_CAPTIVE_PORTAL_DETECTION)
+  // When detection is disabled, captive portal service always sends
+  // RESULT_INTERNET_CONNECTED. Ignore any probe results in that case.
+  if (!captive_portal_detection_enabled_)
+    return;
+  if (type == chrome::NOTIFICATION_CAPTIVE_PORTAL_CHECK_RESULT) {
+    captive_portal_probe_completed_ = true;
+    CaptivePortalService::Results* results =
+        content::Details<CaptivePortalService::Results>(
+            details).ptr();
+    // If a captive portal was detected at any point when the interstitial was
+    // displayed, assume that the interstitial was caused by a captive portal.
+    // Example scenario:
+    // 1- Interstitial displayed and captive portal detected, setting the flag.
+    // 2- Captive portal detection automatically opens portal login page.
+    // 3- User logs in on the portal login page.
+    // A notification will be received here for RESULT_INTERNET_CONNECTED. Make
+    // sure we don't clear the captive protal flag, since the interstitial was
+    // potentially caused by the captive portal.
+    captive_portal_detected_ = captive_portal_detected_ ||
+        (results->result == captive_portal::RESULT_BEHIND_CAPTIVE_PORTAL);
+    // Also keep track of non-HTTP portals and error cases.
+    captive_portal_no_response_ = captive_portal_no_response_ ||
+        (results->result == captive_portal::RESULT_NO_RESPONSE);
+  }
+#endif
+}