Find reasons for the SSL common name invalid error.

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 ae9283c190f73594e2622102a2336d99b318f4c2..47e40968f198ceb9baa041c64e7655affc6f9caf 100644
--- a/chrome/browser/ssl/ssl_error_classification.cc
+++ b/chrome/browser/ssl/ssl_error_classification.cc
@@ -2,15 +2,22 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
+#include <vector>
+
 #include "chrome/browser/ssl/ssl_error_classification.h"
 
 #include "base/build_time.h"
 #include "base/metrics/field_trial.h"
 #include "base/metrics/histogram.h"
+#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 "components/network_time/network_time_tracker.h"
+#include "chrome/browser/ssl/ssl_error_info.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;
@@ -22,9 +29,18 @@ namespace {
 enum SSLInterstitialCause {
   CLOCK_PAST,
   CLOCK_FUTURE,
+  WWW_SUBDOMAIN_MATCH,
+  SUBDOMAIN_MATCH,
+  SUBDOMAIN_INVERSE_MATCH,
+  SUBDOMAIN_OUTSIDE_WILDCARD,
+  SELF_SIGNED,
   UNUSED_INTERSTITIAL_CAUSE_ENTRY,
 };
 
+// Scores/weights which will be constant through all the SSL error types.
+static const float kServerWeight = 0.5f;
+static const float kClientWeight = 0.5f;
+
 void RecordSSLInterstitialCause(bool overridable, SSLInterstitialCause event) {
   if (overridable) {
     UMA_HISTOGRAM_ENUMERATION("interstitial.ssl.cause.overridable", event,
@@ -35,36 +51,62 @@ void RecordSSLInterstitialCause(bool overridable, SSLInterstitialCause event) {
   }
 }
 
+// Utility function - For two unequal strings which have been tokenized, this
+// method checks to see whether |tokenized_potential_subdomain| is a subdomain
+// of |tokenized_parent| and if it is then it returns the difference in the
+// number of tokens between both the vectors, i.e. the difference in the vector
+// size.
+size_t FindSubDomainDifference(
+    const std::vector<std::string>& tokenized_potential_subdomain,
+    const std::vector<std::string>& tokenized_parent) {
+  // A check to ensure that the number of tokens in the tokenized_parent is
+  // less than the tokenized_potential_subdomain.
+  if (tokenized_parent.size() >= tokenized_potential_subdomain.size())
+    return 0;
+
+  size_t tokens_match = 0;
+  size_t diff_size = tokenized_potential_subdomain.size() -
+      tokenized_parent.size();
+  for (size_t i = 0; i < tokenized_parent.size(); ++i) {
+    if (tokenized_parent[i] == tokenized_potential_subdomain[i + diff_size])
+      tokens_match++;
+  }
+  if (tokens_match == tokenized_parent.size())
+    return diff_size;
+  return 0;
+}
+
 } // namespace
 
 SSLErrorClassification::SSLErrorClassification(
     base::Time current_time,
+    const GURL& url,
     const net::X509Certificate& cert)
   : current_time_(current_time),
+    request_url_(url),
     cert_(cert) { }
 
 SSLErrorClassification::~SSLErrorClassification() { }
 
-float SSLErrorClassification::InvalidDateSeverityScore() const {
-  // Client-side characterisitics. Check whether the system's clock is wrong or
-  // not and whether the user has encountered this error before or not.
+float SSLErrorClassification::InvalidDateSeverityScore() const{
+  // 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.
   float severity_date_score = 0.0f;
 
-  static const float kClientWeight = 0.5f;
+  static const float kCertificateExpiredWeight = 0.3f;
+  static const float kNotYetValidWeight = 0.2f;
+
   static const float kSystemClockWeight = 0.75f;
   static const float kSystemClockWrongWeight = 0.1f;
   static const float kSystemClockRightWeight = 1.0f;
 
-  static const float kServerWeight = 0.5f;
-  static const float kCertificateExpiredWeight = 0.3f;
-  static const float kNotYetValidWeight = 0.2f;
-
   if (IsUserClockInThePast(current_time_)  ||
       IsUserClockInTheFuture(current_time_)) {
-    severity_date_score = kClientWeight * kSystemClockWeight *
+    severity_date_score += kClientWeight * kSystemClockWeight *
         kSystemClockWrongWeight;
   } else {
-    severity_date_score = kClientWeight * kSystemClockWeight *
+    severity_date_score += kClientWeight * kSystemClockWeight *
         kSystemClockRightWeight;
   }
   // TODO(radhikabhar): (crbug.com/393262) Check website settings.
@@ -81,6 +123,78 @@ float SSLErrorClassification::InvalidDateSeverityScore() const {
   return severity_date_score;
 }
 
+float SSLErrorClassification::InvalidCommonNameSeverityScore() const {
+  float severity_name_score = 0.0f;
+
+  static const float kWWWDifferenceWeight = 0.3f;
+  static const float kSubDomainWeight = 0.2f;
+  static const float kSubDomainInverseWeight = 1.0f;
+
+  std::string host_name = request_url_.host();
+  if (IsHostNameKnownTLD(host_name)) {
+    std::vector<std::string> host_name_tokens;
+    base::SplitStringDontTrim(host_name,
+                              '.',
+                              &host_name_tokens);
+    if (IsWWWSubDomainMatch())
+      severity_name_score += kServerWeight * kWWWDifferenceWeight;
+    if (IsSubDomainOutsideWildcard(host_name_tokens))
+      severity_name_score += kServerWeight * kWWWDifferenceWeight;
+
+    std::vector<std::string> dns_names;
+    cert_.GetDNSNames(&dns_names);
+    std::vector<std::vector<std::string>> dns_name_tokens =
+        GetTokenizedDNSNames(dns_names);
+    if (IsSubDomainMatch(host_name_tokens, dns_name_tokens))
+      severity_name_score += kServerWeight * kSubDomainWeight;
+    // Inverse case is more likely to be a MITM attack.
+    if (IsSubDomainInverseMatch(host_name_tokens, dns_name_tokens))
+      severity_name_score += kServerWeight * kSubDomainInverseWeight;
+  }
+  return severity_name_score;
+}
+
+void SSLErrorClassification::RecordUMAStatistics(bool overridable,
+                                                 int cert_error) {
+  SSLErrorInfo::ErrorType type =
+      SSLErrorInfo::NetErrorToErrorType(cert_error);
+
+  if (type == SSLErrorInfo::CERT_DATE_INVALID) {
+    if (IsUserClockInThePast(base::Time::NowFromSystemTime()))
+      RecordSSLInterstitialCause(overridable, CLOCK_PAST);
+    if (IsUserClockInTheFuture(base::Time::NowFromSystemTime()))
+      RecordSSLInterstitialCause(overridable, CLOCK_FUTURE);
+  }
+
+  if (type == SSLErrorInfo::CERT_COMMON_NAME_INVALID) {
+    std::string host_name = request_url_.host();
+    if (IsHostNameKnownTLD(host_name)) {

[felt, 2014/07/21 21:34:13]

Suggestion: you might want to record how often the IsHostNameKnownTLD check
fails here, with an UMA stat.

[radhikabhar, 2014/07/22 16:03:38]

Done.

+      std::vector<std::string> host_name_tokens;
+      base::SplitStringDontTrim(host_name,
+                                '.',
+                                &host_name_tokens);
+      if (IsWWWSubDomainMatch())
+        RecordSSLInterstitialCause(overridable, WWW_SUBDOMAIN_MATCH);
+      if (IsSubDomainOutsideWildcard(host_name_tokens))
+        RecordSSLInterstitialCause(overridable, SUBDOMAIN_OUTSIDE_WILDCARD);
+
+      std::vector<std::string> dns_names;
+      cert_.GetDNSNames(&dns_names);
+      std::vector<std::vector<std::string>> dns_name_tokens =
+          GetTokenizedDNSNames(dns_names);
+      if (IsSubDomainMatch(host_name_tokens, dns_name_tokens))
+        RecordSSLInterstitialCause(overridable, SUBDOMAIN_MATCH);
+      if (IsSubDomainInverseMatch(host_name_tokens, dns_name_tokens))
+        RecordSSLInterstitialCause(overridable, SUBDOMAIN_INVERSE_MATCH);
+    }
+  }
+
+  if (type == SSLErrorInfo::CERT_AUTHORITY_INVALID) {
+    if (IsSelfSigned())
+      RecordSSLInterstitialCause(overridable, SELF_SIGNED);
+  }
+}
+
 base::TimeDelta SSLErrorClassification::TimePassedSinceExpiry() const {
   base::TimeDelta delta = current_time_ - cert_.valid_expiry();
   return delta;
@@ -116,9 +230,149 @@ bool SSLErrorClassification::IsUserClockInTheFuture(base::Time time_now) {
   return false;
 }
 
-void SSLErrorClassification::RecordUMAStatistics(bool overridable) {
-  if (IsUserClockInThePast(base::Time::NowFromSystemTime()))
-    RecordSSLInterstitialCause(overridable, CLOCK_PAST);
-  if (IsUserClockInTheFuture(base::Time::NowFromSystemTime()))
-    RecordSSLInterstitialCause(overridable, CLOCK_FUTURE);
+bool SSLErrorClassification::IsHostNameKnownTLD(const std::string& host_name) {
+  size_t tld_length =
+      net::registry_controlled_domains::GetRegistryLength(
+          host_name,
+          net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES,
+          net::registry_controlled_domains::INCLUDE_PRIVATE_REGISTRIES);
+  if (tld_length == 0 || tld_length == std::string::npos)
+    return false;
+  return true;
+}
+
+std::vector<std::vector<std::string>> SSLErrorClassification::
+GetTokenizedDNSNames(std::vector<std::string>& dns_names) const{
+  std::vector<std::vector<std::string>> dns_name_tokens;
+  for (size_t i = 0; i < dns_names.size(); ++i) {
+    std::vector<std::string> dns_name_token_single;
+    if (dns_names[i].empty() || dns_names[i].find('\0') != std::string::npos
+        || !(IsHostNameKnownTLD(dns_names[i]))) {
+      dns_name_token_single.push_back(std::string());
+    } else {
+      base::SplitStringDontTrim(dns_names[i],
+                                '.',
+                                &dns_name_token_single);
+    }
+    dns_name_tokens.push_back(dns_name_token_single);
+  }
+  return dns_name_tokens;
+}
+
+// We accept the inverse case for www for historical reasons.
+bool SSLErrorClassification::IsWWWSubDomainMatch() const {
+  std::string host_name = request_url_.host();
+  if (IsHostNameKnownTLD(host_name)) {
+    std::vector<std::string> dns_names;
+    cert_.GetDNSNames(&dns_names);
+    bool result = false;
+    // Need to account for all possible domains given in the SSL certificate.
+    for (size_t i = 0; i < dns_names.size(); ++i) {
+      if (dns_names[i].empty() || dns_names[i].find('\0') != std::string::npos
+          || dns_names[i].length() == host_name.length()
+          || !(IsHostNameKnownTLD(dns_names[i]))) {
+        result = result || false;
+      } else if (dns_names[i].length() > host_name.length()) {
+        result = result ||
+            net::StripWWW(base::ASCIIToUTF16(dns_names[i])) ==
+            base::ASCIIToUTF16(host_name);
+      } else {
+        result = result ||
+            net::StripWWW(base::ASCIIToUTF16(host_name)) ==
+            base::ASCIIToUTF16(dns_names[i]);
+      }
+    }
+    return result;
+  }
+  return false;
+}
+
+bool SSLErrorClassification::IsSubDomainMatch(
+    const std::vector<std::string>& host_name_tokens,
+    const std::vector<std::vector<std::string>>& dns_name_tokens) const {
+  bool result = false;
+  // Need to account for all the possible domains given in the SSL certificate.
+  for (size_t i = 0; i < dns_name_tokens.size(); ++i) {
+    if (dns_name_tokens[i].empty() ||
+        dns_name_tokens[i].size() >= host_name_tokens.size()) {
+      result = result || false;
+    } else {
+      size_t domain_diff = FindSubDomainDifference(host_name_tokens,
+                                                   dns_name_tokens[i]);
+      if (domain_diff == 1 &&  host_name_tokens[0] != "www")
+        result = result || true;
+    }
+  }
+  return result;
+}
+
+// The inverse case should be treated carefully as this is most likely a MITM
+// attack. We don't want foo.appspot.com to be able to MITM for appspot.com.
+bool SSLErrorClassification::IsSubDomainInverseMatch(
+    const std::vector<std::string>& host_name_tokens,
+    const std::vector<std::vector<std::string>>& dns_name_tokens) const {
+  bool result = false;
+  // Need to account for all the possible domains given in the SSL certificate.
+  for (size_t i = 0; i < dns_name_tokens.size(); ++i) {
+    if (dns_name_tokens[i].empty() ||
+        dns_name_tokens[i].size() <= host_name_tokens.size()) {
+      result = result || false;
+    } else {
+      size_t domain_diff = FindSubDomainDifference(dns_name_tokens[i],
+                                                   host_name_tokens);
+      if (domain_diff == 1 &&  dns_name_tokens[i][0] != "www")
+        result = result || true;
+    }
+  }
+  return result;
+}
+
+// This method is valid for wildcard certificates only.
+bool SSLErrorClassification::IsSubDomainOutsideWildcard(
+    const std::vector<std::string>& host_name_tokens) const {
+  std::string host_name = request_url_.host();
+  std::vector<std::string> dns_names;
+  cert_.GetDNSNames(&dns_names);
+  bool result = false;
+
+  // This method requires that the host name be longer than the dns name on
+  // the certificate.
+  for (size_t i = 0; i < dns_names.size(); ++i) {
+    if (!(dns_names[i][0] == '*' && dns_names[i][1] == '.')) {
+      result = result || false;
+    } else {
+      if (dns_names[i].empty() || dns_names[i].find('\0') != std::string::npos
+          || dns_names[i].length() >= host_name.length()
+          || !(IsHostNameKnownTLD(dns_names[i]))) {
+        result = result || false;
+      } else {
+        // Move past the '*.'.
+        std::string extracted_dns_name = dns_names[i].substr(2);
+        std::vector<std::string> extracted_dns_name_tokens;
+        base::SplitStringDontTrim(extracted_dns_name,
+                                  '.',
+                                  &extracted_dns_name_tokens);
+        size_t domain_diff = FindSubDomainDifference(host_name_tokens,
+                                                     extracted_dns_name_tokens);
+        if (domain_diff == 2)
+          result = result || true;
+      }
+    }
+  }
+  return result;
+}
+
+bool SSLErrorClassification::IsSelfSigned() const {
+  // Check whether the issuer and the subject are the same.
+  const net::CertPrincipal& subject = cert_.subject();
+  const net::CertPrincipal& issuer = cert_.issuer();
+  bool result = subject.common_name == issuer.common_name &&
+      subject.locality_name == issuer.locality_name &&
+      subject.state_or_province_name == issuer.state_or_province_name &&
+      subject.country_name == issuer.country_name &&
+      subject.street_addresses == issuer.street_addresses &&
+      subject.organization_names == issuer.organization_names &&
+      subject.organization_unit_names == issuer.organization_unit_names &&
+      subject.domain_components == issuer.domain_components;
+  return result;
 }