PasswordProtectionService verdict cache management

components/safe_browsing/password_protection/password_protection_service.cc

 // Copyright 2017 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "components/safe_browsing/password_protection/password_protection_service.h"
 
 #include "base/bind.h"
 #include "base/callback.h"
+#include "base/memory/ptr_util.h"
 #include "base/metrics/histogram_macros.h"
+#include "base/strings/string_number_conversions.h"
+#include "base/strings/string_split.h"
+#include "base/strings/string_util.h"
 #include "components/safe_browsing_db/database_manager.h"
+#include "components/safe_browsing_db/v4_protocol_manager_util.h"
 #include "content/public/browser/browser_thread.h"
 
 using content::BrowserThread;
 
 namespace safe_browsing {
 
+namespace {
+
+// Keys for storing password protection verdict into a DictionaryValue.
+static const char kCacheTTL[] = "cache_ttl";
+static const char kCacheCreationTime[] = "cache_creation_time";
+static const char kVerdictType[] = "verdict";
+
+// Helper function to determine if the given origin matches content settings
+// map's patterns.
+bool OriginMatchPrimaryPattern(
+    const GURL& origin,
+    const ContentSettingsPattern& primary_pattern,
+    const ContentSettingsPattern& secondary_pattern_unused) {
+  return ContentSettingsPattern::FromURLNoWildcard(origin) == primary_pattern;
+}
+
+// Convert a LoginReputationClientResponse proto into a DictionaryValue.
+std::unique_ptr<base::DictionaryValue> CreateDictionaryFromVerdict(
+    const LoginReputationClientResponse* verdict,
+    const base::Time& receive_time) {
+  std::unique_ptr<base::DictionaryValue> result =
+      base::MakeUnique<base::DictionaryValue>();
+  if (verdict) {
+    result->SetIntegerWithoutPathExpansion(kCacheTTL,
+                                           verdict->cache_duration_sec());
+    result->SetIntegerWithoutPathExpansion(
+        kCacheCreationTime, static_cast<int>(receive_time.ToDoubleT()));
+    result->SetIntegerWithoutPathExpansion(kVerdictType,
+                                           verdict->verdict_type());
+  }
+  return result;
+}
+
+// Returns the number of path segments |cache_expression| has.
+// For example, return 0 for "www.123.com", since there is no path after the
+// origin; return 3 for www.123.com/abc/def/gh.html since there are three
+// path segments after its origin.
+size_t GetPathDepth(const base::StringPiece& cache_expression) {
+  return base::SplitString(cache_expression, "/", base::KEEP_WHITESPACE,
+                           base::SPLIT_WANT_NONEMPTY)
+      .size();
+}
+
+}  // namespace
+
 PasswordProtectionService::PasswordProtectionService(
     const scoped_refptr<SafeBrowsingDatabaseManager>& database_manager)
     : database_manager_(database_manager), weak_factory_(this) {
   DCHECK_CURRENTLY_ON(BrowserThread::UI);
 }
 
 PasswordProtectionService::~PasswordProtectionService() {
   weak_factory_.InvalidateWeakPtrs();
 }
 
 void PasswordProtectionService::RecordPasswordReuse(const GURL& url) {
   DCHECK_CURRENTLY_ON(BrowserThread::UI);
   DCHECK(database_manager_);
   if (!url.is_valid())
     return;
 
   BrowserThread::PostTaskAndReplyWithResult(
       BrowserThread::IO, FROM_HERE,
       base::Bind(&SafeBrowsingDatabaseManager::MatchCsdWhitelistUrl,
                  database_manager_, url),
       base::Bind(&PasswordProtectionService::OnMatchCsdWhiteListResult,
                  GetWeakPtr()));
 }
 
+LoginReputationClientResponse::VerdictType
+PasswordProtectionService::GetCachedVerdict(
+    const HostContentSettingsMap* settings,
+    const GURL& url) {
+  // TODO(jialiul): Add UMA metrics to track if verdict is available, not
+  // available, or expired.
+  DCHECK(settings);
+  if (!url.is_valid()) {
+    return LoginReputationClientResponse::VERDICT_TYPE_UNSPECIFIED;
+  }
+
+  GURL origin_url = url.GetOrigin();
+  std::unique_ptr<base::DictionaryValue> verdict_dictionary =
+      base::DictionaryValue::From(settings->GetWebsiteSetting(
+          origin_url, GURL(), CONTENT_SETTINGS_TYPE_PASSWORD_PROTECTION,
+          std::string(), nullptr));
+  // Return early if there is no verdict cached for this |origin_url|.
+  if (!verdict_dictionary.get() || verdict_dictionary->empty()) {
+    return LoginReputationClientResponse::VERDICT_TYPE_UNSPECIFIED;
+  }
+
+  // For all the verdicts of the same origin, we key them by cache_expression
+  // field. And its corresponding value is a DictionaryValue contains its ttl,
+  // creation time, and verdict type info.
+  std::vector<std::string> hosts;
+  std::vector<std::string> paths;
+  V4ProtocolManagerUtil::GenerateHostAndPathVariants(url, &hosts, &paths);
+  int cache_creation_time;
+  int cache_duration;
+  int verdict_type_value;
+  base::DictionaryValue* verdict_entry = nullptr;
+  int max_path_depth = 0;
+  LoginReputationClientResponse::VerdictType most_matching_verdict;
+  for (base::DictionaryValue::Iterator it(*verdict_dictionary.get());
+       !it.IsAtEnd(); it.Advance()) {
+    base::StringPiece cache_expression_trimed(
+        base::TrimString(it.key(), "/", base::TrimPositions::TRIM_TRAILING));
+    int path_depth = GetPathDepth(cache_expression_trimed);
+    if (path_depth <= max_path_depth)
+      continue;
+    DCHECK(verdict_dictionary->GetDictionaryWithoutPathExpansion(
+        it.key(), &verdict_entry));
+    DCHECK(ParseVerdictEntry(verdict_entry, &cache_creation_time,
+                             &cache_duration, &verdict_type_value));
+
+    if (HostPathVariantsMatchCacheExpression(
+            hosts, paths, cache_expression_trimed /* cache_expression */) &&
+        !IsCacheExpired(cache_creation_time, cache_duration)) {
+      max_path_depth = path_depth;
+      most_matching_verdict =
+          static_cast<LoginReputationClientResponse::VerdictType>(
+              verdict_type_value);
+    }
+  }
+  return most_matching_verdict;
+}
+
+void PasswordProtectionService::CacheVerdict(
+    const GURL& url,
+    LoginReputationClientResponse* verdict,
+    const base::Time& receive_time,
+    HostContentSettingsMap* settings) {
+  DCHECK(verdict);
+
+  GURL origin_url = url.GetOrigin();
+  std::unique_ptr<base::DictionaryValue> verdict_dictionary =
+      base::DictionaryValue::From(settings->GetWebsiteSetting(
+          origin_url, GURL(), CONTENT_SETTINGS_TYPE_PASSWORD_PROTECTION,
+          std::string(), nullptr));
+
+  if (!verdict_dictionary.get())
+    verdict_dictionary = base::MakeUnique<base::DictionaryValue>();
+
+  std::unique_ptr<base::DictionaryValue> verdict_entry =
+      CreateDictionaryFromVerdict(verdict, receive_time);
+  // If same cache_expression is already in this verdict_dictionary, we simply
+  // override it with a fresher verdict.
+  verdict_dictionary->SetWithoutPathExpansion(verdict->cache_expression(),
+                                              std::move(verdict_entry));
+  settings->SetWebsiteSettingDefaultScope(
+      origin_url, GURL(), CONTENT_SETTINGS_TYPE_PASSWORD_PROTECTION,
+      std::string(), std::move(verdict_dictionary));
+}
+
 void PasswordProtectionService::OnMatchCsdWhiteListResult(
     bool match_whitelist) {
   DCHECK_CURRENTLY_ON(BrowserThread::UI);
   UMA_HISTOGRAM_BOOLEAN(
       "PasswordManager.PasswordReuse.MainFrameMatchCsdWhitelist",
       match_whitelist);
 }
 
+HostContentSettingsMap*
+PasswordProtectionService::GetSettingMapForActiveProfile() {
+  return nullptr;
+}
+
+void PasswordProtectionService::OnURLsDeleted(
+    history::HistoryService* history_service,
+    bool all_history,
+    bool expired,
+    const history::URLRows& deleted_rows,
+    const std::set<GURL>& favicon_urls) {
+  HostContentSettingsMap* setting_map = GetSettingMapForActiveProfile();
+  if (!setting_map)
+    return;
+
+  BrowserThread::PostTask(
+      BrowserThread::UI, FROM_HERE,
+      base::Bind(&PasswordProtectionService::RemoveContentSettingsOnURLsDeleted,
+                 GetWeakPtr(), all_history, deleted_rows, setting_map));
+}
+
+void PasswordProtectionService::RemoveContentSettingsOnURLsDeleted(
+    bool all_history,
+    const history::URLRows& deleted_rows,
+    HostContentSettingsMap* setting_map) {
+  DCHECK_CURRENTLY_ON(BrowserThread::UI);
+  if (all_history) {
+    setting_map->ClearSettingsForOneType(
+        CONTENT_SETTINGS_TYPE_PASSWORD_PROTECTION);
+    return;
+  }
+
+  // For now, if a URL is deleted from history, we simply remove all the
+  // cached verdicts of the same origin. This is a pretty aggressive deletion.
+  // We might revisit this logic later to decide if we want to only delete the
+  // cached verdict whose cache expression matches this URL.
+  for (const history::URLRow& row : deleted_rows) {
+    setting_map->ClearSettingsForOneTypeWithPredicate(
+        CONTENT_SETTINGS_TYPE_PASSWORD_PROTECTION,
+        base::Bind(&OriginMatchPrimaryPattern, row.url().GetOrigin()));
+  }
+}
+
+// static
+bool PasswordProtectionService::ParseVerdictEntry(
+    const base::DictionaryValue* verdict_entry,
+    int* out_cache_creation_time,
+    int* out_cache_duration,
+    int* out_verdict_type) {
+  return verdict_entry &&
+         verdict_entry->GetInteger(kCacheCreationTime,
+                                   out_cache_creation_time) &&
+         verdict_entry->GetInteger(kCacheTTL, out_cache_duration) &&
+         verdict_entry->GetInteger(kVerdictType, out_verdict_type);
+}
+
+bool PasswordProtectionService::HostPathVariantsMatchCacheExpression(
+    const std::vector<std::string>& hosts,
+    const std::vector<std::string>& paths,
+    const base::StringPiece& cache_expression) {
+  for (const auto& host : hosts)
+    for (const auto& path : paths) {
+      std::string host_path_combine(host);
+      host_path_combine.append(path);
+      if (cache_expression ==
+          base::TrimString(host_path_combine, "/",
+                           base::TrimPositions::TRIM_TRAILING)) {
+        return true;
+      }
+    }
+  return false;
+}
+
+bool PasswordProtectionService::IsCacheExpired(int cache_creation_time,
+                                               int cache_duration) {
+  // TODO(jialiul): For now, we assume client's clock is accurate or almost
+  // accurate. Need some logic to handle cases where client's clock is way off.
+  return base::Time::Now().ToDoubleT() >
+         static_cast<double>(cache_creation_time + cache_duration);
+}
+
 }  // namespace safe_browsing