NaCl docs: add sanitizers to GSoC ideas
[chromium-blink-merge.git] / chrome / browser / safe_browsing / download_protection_service.cc
blob7269a4fc5e79a34e3e1020ea06660ecc2261a457
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.
5 #include "chrome/browser/safe_browsing/download_protection_service.h"
7 #include "base/bind.h"
8 #include "base/compiler_specific.h"
9 #include "base/format_macros.h"
10 #include "base/memory/scoped_ptr.h"
11 #include "base/memory/weak_ptr.h"
12 #include "base/metrics/field_trial.h"
13 #include "base/metrics/histogram.h"
14 #include "base/metrics/sparse_histogram.h"
15 #include "base/sequenced_task_runner_helpers.h"
16 #include "base/stl_util.h"
17 #include "base/strings/string_number_conversions.h"
18 #include "base/strings/string_util.h"
19 #include "base/strings/stringprintf.h"
20 #include "base/task/cancelable_task_tracker.h"
21 #include "base/threading/sequenced_worker_pool.h"
22 #include "base/time/time.h"
23 #include "chrome/browser/browser_process.h"
24 #include "chrome/browser/history/history_service.h"
25 #include "chrome/browser/history/history_service_factory.h"
26 #include "chrome/browser/profiles/profile.h"
27 #include "chrome/browser/safe_browsing/binary_feature_extractor.h"
28 #include "chrome/browser/safe_browsing/download_feedback_service.h"
29 #include "chrome/browser/safe_browsing/safe_browsing_service.h"
30 #include "chrome/browser/safe_browsing/sandboxed_zip_analyzer.h"
31 #include "chrome/browser/ui/browser.h"
32 #include "chrome/browser/ui/browser_list.h"
33 #include "chrome/common/safe_browsing/csd.pb.h"
34 #include "chrome/common/safe_browsing/download_protection_util.h"
35 #include "chrome/common/safe_browsing/zip_analyzer.h"
36 #include "chrome/common/url_constants.h"
37 #include "components/google/core/browser/google_util.h"
38 #include "content/public/browser/browser_thread.h"
39 #include "content/public/browser/download_item.h"
40 #include "content/public/browser/page_navigator.h"
41 #include "crypto/sha2.h"
42 #include "google_apis/google_api_keys.h"
43 #include "net/base/escape.h"
44 #include "net/base/load_flags.h"
45 #include "net/cert/x509_cert_types.h"
46 #include "net/cert/x509_certificate.h"
47 #include "net/http/http_status_code.h"
48 #include "net/url_request/url_fetcher.h"
49 #include "net/url_request/url_fetcher_delegate.h"
50 #include "net/url_request/url_request_context_getter.h"
51 #include "net/url_request/url_request_status.h"
53 using content::BrowserThread;
55 namespace {
56 static const int64 kDownloadRequestTimeoutMs = 7000;
57 } // namespace
59 namespace safe_browsing {
61 const char DownloadProtectionService::kDownloadRequestUrl[] =
62 "https://sb-ssl.google.com/safebrowsing/clientreport/download";
64 namespace {
65 // List of extensions for which we track some UMA stats.
66 enum MaliciousExtensionType {
67 EXTENSION_EXE,
68 EXTENSION_MSI,
69 EXTENSION_CAB,
70 EXTENSION_SYS,
71 EXTENSION_SCR,
72 EXTENSION_DRV,
73 EXTENSION_BAT,
74 EXTENSION_ZIP,
75 EXTENSION_RAR,
76 EXTENSION_DLL,
77 EXTENSION_PIF,
78 EXTENSION_COM,
79 EXTENSION_JAR,
80 EXTENSION_CLASS,
81 EXTENSION_PDF,
82 EXTENSION_VB,
83 EXTENSION_REG,
84 EXTENSION_GRP,
85 EXTENSION_OTHER, // Groups all other extensions into one bucket.
86 EXTENSION_CRX,
87 EXTENSION_APK,
88 EXTENSION_DMG,
89 EXTENSION_PKG,
90 EXTENSION_TORRENT,
91 EXTENSION_MAX,
94 MaliciousExtensionType GetExtensionType(const base::FilePath& f) {
95 if (f.MatchesExtension(FILE_PATH_LITERAL(".exe"))) return EXTENSION_EXE;
96 if (f.MatchesExtension(FILE_PATH_LITERAL(".msi"))) return EXTENSION_MSI;
97 if (f.MatchesExtension(FILE_PATH_LITERAL(".cab"))) return EXTENSION_CAB;
98 if (f.MatchesExtension(FILE_PATH_LITERAL(".sys"))) return EXTENSION_SYS;
99 if (f.MatchesExtension(FILE_PATH_LITERAL(".scr"))) return EXTENSION_SCR;
100 if (f.MatchesExtension(FILE_PATH_LITERAL(".drv"))) return EXTENSION_DRV;
101 if (f.MatchesExtension(FILE_PATH_LITERAL(".bat"))) return EXTENSION_BAT;
102 if (f.MatchesExtension(FILE_PATH_LITERAL(".zip"))) return EXTENSION_ZIP;
103 if (f.MatchesExtension(FILE_PATH_LITERAL(".rar"))) return EXTENSION_RAR;
104 if (f.MatchesExtension(FILE_PATH_LITERAL(".dll"))) return EXTENSION_DLL;
105 if (f.MatchesExtension(FILE_PATH_LITERAL(".pif"))) return EXTENSION_PIF;
106 if (f.MatchesExtension(FILE_PATH_LITERAL(".com"))) return EXTENSION_COM;
107 if (f.MatchesExtension(FILE_PATH_LITERAL(".jar"))) return EXTENSION_JAR;
108 if (f.MatchesExtension(FILE_PATH_LITERAL(".class"))) return EXTENSION_CLASS;
109 if (f.MatchesExtension(FILE_PATH_LITERAL(".pdf"))) return EXTENSION_PDF;
110 if (f.MatchesExtension(FILE_PATH_LITERAL(".vb"))) return EXTENSION_VB;
111 if (f.MatchesExtension(FILE_PATH_LITERAL(".reg"))) return EXTENSION_REG;
112 if (f.MatchesExtension(FILE_PATH_LITERAL(".grp"))) return EXTENSION_GRP;
113 if (f.MatchesExtension(FILE_PATH_LITERAL(".crx"))) return EXTENSION_CRX;
114 if (f.MatchesExtension(FILE_PATH_LITERAL(".apk"))) return EXTENSION_APK;
115 if (f.MatchesExtension(FILE_PATH_LITERAL(".dmg"))) return EXTENSION_DMG;
116 if (f.MatchesExtension(FILE_PATH_LITERAL(".pkg"))) return EXTENSION_PKG;
117 if (f.MatchesExtension(FILE_PATH_LITERAL(".torrent")))
118 return EXTENSION_TORRENT;
119 return EXTENSION_OTHER;
122 void RecordFileExtensionType(const base::FilePath& file) {
123 UMA_HISTOGRAM_ENUMERATION("SBClientDownload.DownloadExtensions",
124 GetExtensionType(file),
125 EXTENSION_MAX);
128 // Enumerate for histogramming purposes.
129 // DO NOT CHANGE THE ORDERING OF THESE VALUES (different histogram data will
130 // be mixed together based on their values).
131 enum SBStatsType {
132 DOWNLOAD_URL_CHECKS_TOTAL,
133 DOWNLOAD_URL_CHECKS_CANCELED,
134 DOWNLOAD_URL_CHECKS_MALWARE,
136 DOWNLOAD_HASH_CHECKS_TOTAL,
137 DOWNLOAD_HASH_CHECKS_MALWARE,
139 // Memory space for histograms is determined by the max.
140 // ALWAYS ADD NEW VALUES BEFORE THIS ONE.
141 DOWNLOAD_CHECKS_MAX
144 // Prepares URLs to be put into a ping message. Currently this just shortens
145 // data: URIs, other URLs are included verbatim.
146 std::string SanitizeUrl(const GURL& url) {
147 std::string spec = url.spec();
148 if (url.SchemeIs(url::kDataScheme)) {
149 size_t comma_pos = spec.find(',');
150 if (comma_pos != std::string::npos && comma_pos != spec.size() - 1) {
151 std::string hash_value = crypto::SHA256HashString(spec);
152 spec.erase(comma_pos + 1);
153 spec += base::HexEncode(hash_value.data(), hash_value.size());
156 return spec;
159 } // namespace
161 // Parent SafeBrowsing::Client class used to lookup the bad binary
162 // URL and digest list. There are two sub-classes (one for each list).
163 class DownloadSBClient
164 : public SafeBrowsingDatabaseManager::Client,
165 public base::RefCountedThreadSafe<DownloadSBClient> {
166 public:
167 DownloadSBClient(
168 const content::DownloadItem& item,
169 const DownloadProtectionService::CheckDownloadCallback& callback,
170 const scoped_refptr<SafeBrowsingUIManager>& ui_manager,
171 SBStatsType total_type,
172 SBStatsType dangerous_type)
173 : sha256_hash_(item.GetHash()),
174 url_chain_(item.GetUrlChain()),
175 referrer_url_(item.GetReferrerUrl()),
176 callback_(callback),
177 ui_manager_(ui_manager),
178 start_time_(base::TimeTicks::Now()),
179 total_type_(total_type),
180 dangerous_type_(dangerous_type) {}
182 virtual void StartCheck() = 0;
183 virtual bool IsDangerous(SBThreatType threat_type) const = 0;
185 protected:
186 friend class base::RefCountedThreadSafe<DownloadSBClient>;
187 ~DownloadSBClient() override {}
189 void CheckDone(SBThreatType threat_type) {
190 DownloadProtectionService::DownloadCheckResult result =
191 IsDangerous(threat_type) ?
192 DownloadProtectionService::DANGEROUS :
193 DownloadProtectionService::SAFE;
194 BrowserThread::PostTask(BrowserThread::UI,
195 FROM_HERE,
196 base::Bind(callback_, result));
197 UpdateDownloadCheckStats(total_type_);
198 if (threat_type != SB_THREAT_TYPE_SAFE) {
199 UpdateDownloadCheckStats(dangerous_type_);
200 BrowserThread::PostTask(
201 BrowserThread::UI,
202 FROM_HERE,
203 base::Bind(&DownloadSBClient::ReportMalware,
204 this, threat_type));
208 void ReportMalware(SBThreatType threat_type) {
209 std::string post_data;
210 if (!sha256_hash_.empty())
211 post_data += base::HexEncode(sha256_hash_.data(),
212 sha256_hash_.size()) + "\n";
213 for (size_t i = 0; i < url_chain_.size(); ++i) {
214 post_data += url_chain_[i].spec() + "\n";
216 ui_manager_->ReportSafeBrowsingHit(
217 url_chain_.back(), // malicious_url
218 url_chain_.front(), // page_url
219 referrer_url_,
220 true, // is_subresource
221 threat_type,
222 post_data);
225 void UpdateDownloadCheckStats(SBStatsType stat_type) {
226 UMA_HISTOGRAM_ENUMERATION("SB2.DownloadChecks",
227 stat_type,
228 DOWNLOAD_CHECKS_MAX);
231 std::string sha256_hash_;
232 std::vector<GURL> url_chain_;
233 GURL referrer_url_;
234 DownloadProtectionService::CheckDownloadCallback callback_;
235 scoped_refptr<SafeBrowsingUIManager> ui_manager_;
236 base::TimeTicks start_time_;
238 private:
239 const SBStatsType total_type_;
240 const SBStatsType dangerous_type_;
242 DISALLOW_COPY_AND_ASSIGN(DownloadSBClient);
245 class DownloadUrlSBClient : public DownloadSBClient {
246 public:
247 DownloadUrlSBClient(
248 const content::DownloadItem& item,
249 const DownloadProtectionService::CheckDownloadCallback& callback,
250 const scoped_refptr<SafeBrowsingUIManager>& ui_manager,
251 const scoped_refptr<SafeBrowsingDatabaseManager>& database_manager)
252 : DownloadSBClient(item, callback, ui_manager,
253 DOWNLOAD_URL_CHECKS_TOTAL,
254 DOWNLOAD_URL_CHECKS_MALWARE),
255 database_manager_(database_manager) { }
257 void StartCheck() override {
258 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
259 if (!database_manager_.get() ||
260 database_manager_->CheckDownloadUrl(url_chain_, this)) {
261 CheckDone(SB_THREAT_TYPE_SAFE);
262 } else {
263 AddRef(); // SafeBrowsingService takes a pointer not a scoped_refptr.
267 bool IsDangerous(SBThreatType threat_type) const override {
268 return threat_type == SB_THREAT_TYPE_BINARY_MALWARE_URL;
271 void OnCheckDownloadUrlResult(const std::vector<GURL>& url_chain,
272 SBThreatType threat_type) override {
273 CheckDone(threat_type);
274 UMA_HISTOGRAM_TIMES("SB2.DownloadUrlCheckDuration",
275 base::TimeTicks::Now() - start_time_);
276 Release();
279 protected:
280 ~DownloadUrlSBClient() override {}
282 private:
283 scoped_refptr<SafeBrowsingDatabaseManager> database_manager_;
285 DISALLOW_COPY_AND_ASSIGN(DownloadUrlSBClient);
288 class DownloadProtectionService::CheckClientDownloadRequest
289 : public base::RefCountedThreadSafe<
290 DownloadProtectionService::CheckClientDownloadRequest,
291 BrowserThread::DeleteOnUIThread>,
292 public net::URLFetcherDelegate,
293 public content::DownloadItem::Observer {
294 public:
295 CheckClientDownloadRequest(
296 content::DownloadItem* item,
297 const CheckDownloadCallback& callback,
298 DownloadProtectionService* service,
299 const scoped_refptr<SafeBrowsingDatabaseManager>& database_manager,
300 BinaryFeatureExtractor* binary_feature_extractor)
301 : item_(item),
302 url_chain_(item->GetUrlChain()),
303 referrer_url_(item->GetReferrerUrl()),
304 tab_url_(item->GetTabUrl()),
305 tab_referrer_url_(item->GetTabReferrerUrl()),
306 zipped_executable_(false),
307 callback_(callback),
308 service_(service),
309 binary_feature_extractor_(binary_feature_extractor),
310 database_manager_(database_manager),
311 pingback_enabled_(service_->enabled()),
312 finished_(false),
313 type_(ClientDownloadRequest::WIN_EXECUTABLE),
314 start_time_(base::TimeTicks::Now()),
315 weakptr_factory_(this) {
316 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
317 item_->AddObserver(this);
320 void Start() {
321 DVLOG(2) << "Starting SafeBrowsing download check for: "
322 << item_->DebugString(true);
323 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
324 // TODO(noelutz): implement some cache to make sure we don't issue the same
325 // request over and over again if a user downloads the same binary multiple
326 // times.
327 DownloadCheckResultReason reason = REASON_MAX;
328 if (!IsSupportedDownload(
329 *item_, item_->GetTargetFilePath(), &reason, &type_)) {
330 switch (reason) {
331 case REASON_EMPTY_URL_CHAIN:
332 case REASON_INVALID_URL:
333 case REASON_UNSUPPORTED_URL_SCHEME:
334 PostFinishTask(UNKNOWN, reason);
335 return;
337 case REASON_NOT_BINARY_FILE:
338 RecordFileExtensionType(item_->GetTargetFilePath());
339 PostFinishTask(UNKNOWN, reason);
340 return;
342 default:
343 // We only expect the reasons explicitly handled above.
344 NOTREACHED();
347 RecordFileExtensionType(item_->GetTargetFilePath());
349 // Compute features from the file contents. Note that we record histograms
350 // based on the result, so this runs regardless of whether the pingbacks
351 // are enabled.
352 if (item_->GetTargetFilePath().MatchesExtension(
353 FILE_PATH_LITERAL(".zip"))) {
354 StartExtractZipFeatures();
355 } else {
356 DCHECK(!download_protection_util::IsArchiveFile(
357 item_->GetTargetFilePath()));
358 StartExtractFileFeatures();
362 // Start a timeout to cancel the request if it takes too long.
363 // This should only be called after we have finished accessing the file.
364 void StartTimeout() {
365 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
366 if (!service_) {
367 // Request has already been cancelled.
368 return;
370 timeout_start_time_ = base::TimeTicks::Now();
371 BrowserThread::PostDelayedTask(
372 BrowserThread::UI,
373 FROM_HERE,
374 base::Bind(&CheckClientDownloadRequest::Cancel,
375 weakptr_factory_.GetWeakPtr()),
376 base::TimeDelta::FromMilliseconds(
377 service_->download_request_timeout_ms()));
380 // Canceling a request will cause us to always report the result as UNKNOWN
381 // unless a pending request is about to call FinishRequest.
382 void Cancel() {
383 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
384 if (fetcher_.get()) {
385 // The DownloadProtectionService is going to release its reference, so we
386 // might be destroyed before the URLFetcher completes. Cancel the
387 // fetcher so it does not try to invoke OnURLFetchComplete.
388 fetcher_.reset();
390 // Note: If there is no fetcher, then some callback is still holding a
391 // reference to this object. We'll eventually wind up in some method on
392 // the UI thread that will call FinishRequest() again. If FinishRequest()
393 // is called a second time, it will be a no-op.
394 FinishRequest(UNKNOWN, REASON_REQUEST_CANCELED);
395 // Calling FinishRequest might delete this object, we may be deleted by
396 // this point.
399 // content::DownloadItem::Observer implementation.
400 void OnDownloadDestroyed(content::DownloadItem* download) override {
401 Cancel();
402 DCHECK(item_ == NULL);
405 // From the net::URLFetcherDelegate interface.
406 void OnURLFetchComplete(const net::URLFetcher* source) override {
407 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
408 DCHECK_EQ(source, fetcher_.get());
409 DVLOG(2) << "Received a response for URL: "
410 << item_->GetUrlChain().back() << ": success="
411 << source->GetStatus().is_success() << " response_code="
412 << source->GetResponseCode();
413 if (source->GetStatus().is_success()) {
414 UMA_HISTOGRAM_SPARSE_SLOWLY(
415 "SBClientDownload.DownloadRequestResponseCode",
416 source->GetResponseCode());
418 UMA_HISTOGRAM_SPARSE_SLOWLY(
419 "SBClientDownload.DownloadRequestNetError",
420 -source->GetStatus().error());
421 DownloadCheckResultReason reason = REASON_SERVER_PING_FAILED;
422 DownloadCheckResult result = UNKNOWN;
423 if (source->GetStatus().is_success() &&
424 net::HTTP_OK == source->GetResponseCode()) {
425 ClientDownloadResponse response;
426 std::string data;
427 bool got_data = source->GetResponseAsString(&data);
428 DCHECK(got_data);
429 if (!response.ParseFromString(data)) {
430 reason = REASON_INVALID_RESPONSE_PROTO;
431 result = UNKNOWN;
432 } else if (response.verdict() == ClientDownloadResponse::SAFE) {
433 reason = REASON_DOWNLOAD_SAFE;
434 result = SAFE;
435 } else if (service_ && !service_->IsSupportedDownload(
436 *item_, item_->GetTargetFilePath())) {
437 // The client of the download protection service assumes that we don't
438 // support this download so we cannot return any other verdict than
439 // UNKNOWN even if the server says it's dangerous to download this file.
440 // Note: if service_ is NULL we already cancelled the request and
441 // returned UNKNOWN.
442 reason = REASON_DOWNLOAD_NOT_SUPPORTED;
443 result = UNKNOWN;
444 } else if (response.verdict() == ClientDownloadResponse::DANGEROUS) {
445 reason = REASON_DOWNLOAD_DANGEROUS;
446 result = DANGEROUS;
447 } else if (response.verdict() == ClientDownloadResponse::UNCOMMON) {
448 reason = REASON_DOWNLOAD_UNCOMMON;
449 result = UNCOMMON;
450 } else if (response.verdict() == ClientDownloadResponse::DANGEROUS_HOST) {
451 reason = REASON_DOWNLOAD_DANGEROUS_HOST;
452 result = DANGEROUS_HOST;
453 } else if (
454 response.verdict() == ClientDownloadResponse::POTENTIALLY_UNWANTED) {
455 reason = REASON_DOWNLOAD_POTENTIALLY_UNWANTED;
456 result = POTENTIALLY_UNWANTED;
457 } else {
458 LOG(DFATAL) << "Unknown download response verdict: "
459 << response.verdict();
460 reason = REASON_INVALID_RESPONSE_VERDICT;
461 result = UNKNOWN;
463 DownloadFeedbackService::MaybeStorePingsForDownload(
464 result, item_, client_download_request_data_, data);
466 // We don't need the fetcher anymore.
467 fetcher_.reset();
468 UMA_HISTOGRAM_TIMES("SBClientDownload.DownloadRequestDuration",
469 base::TimeTicks::Now() - start_time_);
470 UMA_HISTOGRAM_TIMES("SBClientDownload.DownloadRequestNetworkDuration",
471 base::TimeTicks::Now() - request_start_time_);
472 FinishRequest(result, reason);
475 static bool IsSupportedDownload(const content::DownloadItem& item,
476 const base::FilePath& target_path,
477 DownloadCheckResultReason* reason,
478 ClientDownloadRequest::DownloadType* type) {
479 if (item.GetUrlChain().empty()) {
480 *reason = REASON_EMPTY_URL_CHAIN;
481 return false;
483 const GURL& final_url = item.GetUrlChain().back();
484 if (!final_url.is_valid() || final_url.is_empty()) {
485 *reason = REASON_INVALID_URL;
486 return false;
488 if (!download_protection_util::IsBinaryFile(target_path)) {
489 *reason = REASON_NOT_BINARY_FILE;
490 return false;
492 if ((!final_url.IsStandard() && !final_url.SchemeIsBlob() &&
493 !final_url.SchemeIs(url::kDataScheme)) ||
494 final_url.SchemeIsFile()) {
495 *reason = REASON_UNSUPPORTED_URL_SCHEME;
496 return false;
498 *type = download_protection_util::GetDownloadType(target_path);
499 return true;
502 private:
503 friend struct BrowserThread::DeleteOnThread<BrowserThread::UI>;
504 friend class base::DeleteHelper<CheckClientDownloadRequest>;
506 ~CheckClientDownloadRequest() override {
507 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
508 DCHECK(item_ == NULL);
511 void OnFileFeatureExtractionDone() {
512 // This can run in any thread, since it just posts more messages.
514 // TODO(noelutz): DownloadInfo should also contain the IP address of
515 // every URL in the redirect chain. We also should check whether the
516 // download URL is hosted on the internal network.
517 BrowserThread::PostTask(
518 BrowserThread::IO,
519 FROM_HERE,
520 base::Bind(&CheckClientDownloadRequest::CheckWhitelists, this));
522 // We wait until after the file checks finish to start the timeout, as
523 // windows can cause permissions errors if the timeout fired while we were
524 // checking the file signature and we tried to complete the download.
525 BrowserThread::PostTask(
526 BrowserThread::UI,
527 FROM_HERE,
528 base::Bind(&CheckClientDownloadRequest::StartTimeout, this));
531 void StartExtractFileFeatures() {
532 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
533 DCHECK(item_); // Called directly from Start(), item should still exist.
534 // Since we do blocking I/O, offload this to a worker thread.
535 // The task does not need to block shutdown.
536 BrowserThread::GetBlockingPool()->PostWorkerTaskWithShutdownBehavior(
537 FROM_HERE,
538 base::Bind(&CheckClientDownloadRequest::ExtractFileFeatures,
539 this, item_->GetFullPath()),
540 base::SequencedWorkerPool::CONTINUE_ON_SHUTDOWN);
543 void ExtractFileFeatures(const base::FilePath& file_path) {
544 base::TimeTicks start_time = base::TimeTicks::Now();
545 binary_feature_extractor_->CheckSignature(file_path, &signature_info_);
546 bool is_signed = (signature_info_.certificate_chain_size() > 0);
547 if (is_signed) {
548 DVLOG(2) << "Downloaded a signed binary: " << file_path.value();
549 } else {
550 DVLOG(2) << "Downloaded an unsigned binary: "
551 << file_path.value();
553 UMA_HISTOGRAM_BOOLEAN("SBClientDownload.SignedBinaryDownload", is_signed);
554 UMA_HISTOGRAM_TIMES("SBClientDownload.ExtractSignatureFeaturesTime",
555 base::TimeTicks::Now() - start_time);
557 start_time = base::TimeTicks::Now();
558 binary_feature_extractor_->ExtractImageHeaders(file_path, &image_headers_);
559 UMA_HISTOGRAM_TIMES("SBClientDownload.ExtractImageHeadersTime",
560 base::TimeTicks::Now() - start_time);
562 OnFileFeatureExtractionDone();
565 void StartExtractZipFeatures() {
566 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
567 DCHECK(item_); // Called directly from Start(), item should still exist.
568 zip_analysis_start_time_ = base::TimeTicks::Now();
569 // We give the zip analyzer a weak pointer to this object. Since the
570 // analyzer is refcounted, it might outlive the request.
571 analyzer_ = new SandboxedZipAnalyzer(
572 item_->GetFullPath(),
573 base::Bind(&CheckClientDownloadRequest::OnZipAnalysisFinished,
574 weakptr_factory_.GetWeakPtr()));
575 analyzer_->Start();
578 void OnZipAnalysisFinished(const zip_analyzer::Results& results) {
579 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
580 if (!service_)
581 return;
582 if (results.success) {
583 zipped_executable_ = results.has_executable;
584 DVLOG(1) << "Zip analysis finished for " << item_->GetFullPath().value()
585 << ", has_executable=" << results.has_executable
586 << " has_archive=" << results.has_archive;
587 } else {
588 DVLOG(1) << "Zip analysis failed for " << item_->GetFullPath().value();
590 UMA_HISTOGRAM_BOOLEAN("SBClientDownload.ZipFileHasExecutable",
591 zipped_executable_);
592 UMA_HISTOGRAM_BOOLEAN("SBClientDownload.ZipFileHasArchiveButNoExecutable",
593 results.has_archive && !zipped_executable_);
594 UMA_HISTOGRAM_TIMES("SBClientDownload.ExtractZipFeaturesTime",
595 base::TimeTicks::Now() - zip_analysis_start_time_);
597 if (!zipped_executable_) {
598 PostFinishTask(UNKNOWN, REASON_ARCHIVE_WITHOUT_BINARIES);
599 return;
601 OnFileFeatureExtractionDone();
604 static void RecordCountOfSignedOrWhitelistedDownload() {
605 UMA_HISTOGRAM_COUNTS("SBClientDownload.SignedOrWhitelistedDownload", 1);
608 void CheckWhitelists() {
609 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
611 if (!database_manager_.get()) {
612 PostFinishTask(UNKNOWN, REASON_SB_DISABLED);
613 return;
616 const GURL& url = url_chain_.back();
617 if (url.is_valid() && database_manager_->MatchDownloadWhitelistUrl(url)) {
618 DVLOG(2) << url << " is on the download whitelist.";
619 RecordCountOfSignedOrWhitelistedDownload();
620 // TODO(grt): Continue processing without uploading so that
621 // ClientDownloadRequest callbacks can be run even for this type of safe
622 // download.
623 PostFinishTask(SAFE, REASON_WHITELISTED_URL);
624 return;
627 if (signature_info_.trusted()) {
628 RecordCountOfSignedOrWhitelistedDownload();
629 for (int i = 0; i < signature_info_.certificate_chain_size(); ++i) {
630 if (CertificateChainIsWhitelisted(
631 signature_info_.certificate_chain(i))) {
632 // TODO(grt): Continue processing without uploading so that
633 // ClientDownloadRequest callbacks can be run even for this type of
634 // safe download.
635 PostFinishTask(SAFE, REASON_TRUSTED_EXECUTABLE);
636 return;
641 if (!pingback_enabled_) {
642 PostFinishTask(UNKNOWN, REASON_PING_DISABLED);
643 return;
646 // Currently, the UI only works on Windows so we don't even bother with
647 // pinging the server if we're not on Windows.
648 // TODO(noelutz): change this code once the UI is done for Linux and Mac.
649 #if defined(OS_MACOSX)
650 // TODO(mattm): remove this (see crbug.com/414834).
651 if (base::FieldTrialList::FindFullName("SafeBrowsingOSXClientDownloadPings")
652 != "Enabled") {
653 PostFinishTask(UNKNOWN, REASON_OS_NOT_SUPPORTED);
654 return;
656 #endif
657 #if defined(OS_WIN) || defined(OS_MACOSX)
658 // The URLFetcher is owned by the UI thread, so post a message to
659 // start the pingback.
660 BrowserThread::PostTask(
661 BrowserThread::UI,
662 FROM_HERE,
663 base::Bind(&CheckClientDownloadRequest::GetTabRedirects, this));
664 #else
665 PostFinishTask(UNKNOWN, REASON_OS_NOT_SUPPORTED);
666 #endif
669 void GetTabRedirects() {
670 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
671 if (!service_)
672 return;
674 if (!tab_url_.is_valid()) {
675 SendRequest();
676 return;
679 Profile* profile = Profile::FromBrowserContext(item_->GetBrowserContext());
680 HistoryService* history = HistoryServiceFactory::GetForProfile(
681 profile, ServiceAccessType::EXPLICIT_ACCESS);
682 if (!history) {
683 SendRequest();
684 return;
687 history->QueryRedirectsTo(
688 tab_url_,
689 base::Bind(&CheckClientDownloadRequest::OnGotTabRedirects,
690 base::Unretained(this),
691 tab_url_),
692 &request_tracker_);
695 void OnGotTabRedirects(const GURL& url,
696 const history::RedirectList* redirect_list) {
697 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
698 DCHECK_EQ(url, tab_url_);
699 if (!service_)
700 return;
702 if (!redirect_list->empty()) {
703 tab_redirects_.insert(
704 tab_redirects_.end(), redirect_list->rbegin(), redirect_list->rend());
707 SendRequest();
710 void SendRequest() {
711 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
713 // This is our last chance to check whether the request has been canceled
714 // before sending it.
715 if (!service_)
716 return;
718 ClientDownloadRequest request;
719 request.set_url(SanitizeUrl(item_->GetUrlChain().back()));
720 request.mutable_digests()->set_sha256(item_->GetHash());
721 request.set_length(item_->GetReceivedBytes());
722 for (size_t i = 0; i < item_->GetUrlChain().size(); ++i) {
723 ClientDownloadRequest::Resource* resource = request.add_resources();
724 resource->set_url(SanitizeUrl(item_->GetUrlChain()[i]));
725 if (i == item_->GetUrlChain().size() - 1) {
726 // The last URL in the chain is the download URL.
727 resource->set_type(ClientDownloadRequest::DOWNLOAD_URL);
728 resource->set_referrer(SanitizeUrl(item_->GetReferrerUrl()));
729 DVLOG(2) << "dl url " << resource->url();
730 if (!item_->GetRemoteAddress().empty()) {
731 resource->set_remote_ip(item_->GetRemoteAddress());
732 DVLOG(2) << " dl url remote addr: " << resource->remote_ip();
734 DVLOG(2) << "dl referrer " << resource->referrer();
735 } else {
736 DVLOG(2) << "dl redirect " << i << " " << resource->url();
737 resource->set_type(ClientDownloadRequest::DOWNLOAD_REDIRECT);
739 // TODO(noelutz): fill out the remote IP addresses.
741 // TODO(mattm): fill out the remote IP addresses for tab resources.
742 for (size_t i = 0; i < tab_redirects_.size(); ++i) {
743 ClientDownloadRequest::Resource* resource = request.add_resources();
744 DVLOG(2) << "tab redirect " << i << " " << tab_redirects_[i].spec();
745 resource->set_url(SanitizeUrl(tab_redirects_[i]));
746 resource->set_type(ClientDownloadRequest::TAB_REDIRECT);
748 if (tab_url_.is_valid()) {
749 ClientDownloadRequest::Resource* resource = request.add_resources();
750 resource->set_url(SanitizeUrl(tab_url_));
751 DVLOG(2) << "tab url " << resource->url();
752 resource->set_type(ClientDownloadRequest::TAB_URL);
753 if (tab_referrer_url_.is_valid()) {
754 resource->set_referrer(SanitizeUrl(tab_referrer_url_));
755 DVLOG(2) << "tab referrer " << resource->referrer();
759 request.set_user_initiated(item_->HasUserGesture());
760 request.set_file_basename(
761 item_->GetTargetFilePath().BaseName().AsUTF8Unsafe());
762 request.set_download_type(type_);
763 request.mutable_signature()->CopyFrom(signature_info_);
764 request.mutable_image_headers()->CopyFrom(image_headers_);
765 if (!request.SerializeToString(&client_download_request_data_)) {
766 FinishRequest(UNKNOWN, REASON_INVALID_REQUEST_PROTO);
767 return;
770 service_->client_download_request_callbacks_.Notify(item_, &request);
772 DVLOG(2) << "Sending a request for URL: "
773 << item_->GetUrlChain().back();
774 fetcher_.reset(net::URLFetcher::Create(0 /* ID used for testing */,
775 GetDownloadRequestUrl(),
776 net::URLFetcher::POST,
777 this));
778 fetcher_->SetLoadFlags(net::LOAD_DISABLE_CACHE);
779 fetcher_->SetAutomaticallyRetryOn5xx(false); // Don't retry on error.
780 fetcher_->SetRequestContext(service_->request_context_getter_.get());
781 fetcher_->SetUploadData("application/octet-stream",
782 client_download_request_data_);
783 request_start_time_ = base::TimeTicks::Now();
784 UMA_HISTOGRAM_COUNTS("SBClientDownload.DownloadRequestPayloadSize",
785 client_download_request_data_.size());
786 fetcher_->Start();
789 void PostFinishTask(DownloadCheckResult result,
790 DownloadCheckResultReason reason) {
791 BrowserThread::PostTask(
792 BrowserThread::UI,
793 FROM_HERE,
794 base::Bind(&CheckClientDownloadRequest::FinishRequest, this, result,
795 reason));
798 void FinishRequest(DownloadCheckResult result,
799 DownloadCheckResultReason reason) {
800 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
801 if (finished_) {
802 return;
804 finished_ = true;
805 // Ensure the timeout task is cancelled while we still have a non-zero
806 // refcount. (crbug.com/240449)
807 weakptr_factory_.InvalidateWeakPtrs();
808 if (!request_start_time_.is_null()) {
809 UMA_HISTOGRAM_ENUMERATION("SBClientDownload.DownloadRequestNetworkStats",
810 reason,
811 REASON_MAX);
813 if (!timeout_start_time_.is_null()) {
814 UMA_HISTOGRAM_ENUMERATION("SBClientDownload.DownloadRequestTimeoutStats",
815 reason,
816 REASON_MAX);
817 if (reason != REASON_REQUEST_CANCELED) {
818 UMA_HISTOGRAM_TIMES("SBClientDownload.DownloadRequestTimeoutDuration",
819 base::TimeTicks::Now() - timeout_start_time_);
822 if (result == SAFE && (reason == REASON_WHITELISTED_URL ||
823 reason == REASON_TRUSTED_EXECUTABLE)) {
824 // Due to the short-circuit logic in CheckWhitelists (see TODOs there), a
825 // ClientDownloadRequest was not generated for this download and callbacks
826 // were not run. Run them now with null to indicate that a download has
827 // taken place.
828 // TODO(grt): persist metadata for these downloads as well.
829 service_->client_download_request_callbacks_.Notify(item_, nullptr);
831 if (service_) {
832 DVLOG(2) << "SafeBrowsing download verdict for: "
833 << item_->DebugString(true) << " verdict:" << reason
834 << " result:" << result;
835 UMA_HISTOGRAM_ENUMERATION("SBClientDownload.CheckDownloadStats",
836 reason,
837 REASON_MAX);
838 #if defined(OS_MACOSX)
839 // OSX is currently sending pings only for evaluation purposes, ignore
840 // the result for now.
841 // TODO(mattm): remove this and update the ifdef in
842 // DownloadItemImpl::IsDangerous (see crbug.com/413968).
843 result = UNKNOWN;
844 #endif
845 callback_.Run(result);
846 item_->RemoveObserver(this);
847 item_ = NULL;
848 DownloadProtectionService* service = service_;
849 service_ = NULL;
850 service->RequestFinished(this);
851 // DownloadProtectionService::RequestFinished will decrement our refcount,
852 // so we may be deleted now.
853 } else {
854 callback_.Run(UNKNOWN);
858 bool CertificateChainIsWhitelisted(
859 const ClientDownloadRequest_CertificateChain& chain) {
860 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
861 if (chain.element_size() < 2) {
862 // We need to have both a signing certificate and its issuer certificate
863 // present to construct a whitelist entry.
864 return false;
866 scoped_refptr<net::X509Certificate> cert =
867 net::X509Certificate::CreateFromBytes(
868 chain.element(0).certificate().data(),
869 chain.element(0).certificate().size());
870 if (!cert.get()) {
871 return false;
874 for (int i = 1; i < chain.element_size(); ++i) {
875 scoped_refptr<net::X509Certificate> issuer =
876 net::X509Certificate::CreateFromBytes(
877 chain.element(i).certificate().data(),
878 chain.element(i).certificate().size());
879 if (!issuer.get()) {
880 return false;
882 std::vector<std::string> whitelist_strings;
883 DownloadProtectionService::GetCertificateWhitelistStrings(
884 *cert.get(), *issuer.get(), &whitelist_strings);
885 for (size_t j = 0; j < whitelist_strings.size(); ++j) {
886 if (database_manager_->MatchDownloadWhitelistString(
887 whitelist_strings[j])) {
888 DVLOG(2) << "Certificate matched whitelist, cert="
889 << cert->subject().GetDisplayName()
890 << " issuer=" << issuer->subject().GetDisplayName();
891 return true;
894 cert = issuer;
896 return false;
899 // The DownloadItem we are checking. Will be NULL if the request has been
900 // canceled. Must be accessed only on UI thread.
901 content::DownloadItem* item_;
902 // Copies of data from |item_| for access on other threads.
903 std::vector<GURL> url_chain_;
904 GURL referrer_url_;
905 // URL chain of redirects leading to (but not including) |tab_url|.
906 std::vector<GURL> tab_redirects_;
907 // URL and referrer of the window the download was started from.
908 GURL tab_url_;
909 GURL tab_referrer_url_;
911 bool zipped_executable_;
912 ClientDownloadRequest_SignatureInfo signature_info_;
913 ClientDownloadRequest_ImageHeaders image_headers_;
914 CheckDownloadCallback callback_;
915 // Will be NULL if the request has been canceled.
916 DownloadProtectionService* service_;
917 scoped_refptr<BinaryFeatureExtractor> binary_feature_extractor_;
918 scoped_refptr<SafeBrowsingDatabaseManager> database_manager_;
919 const bool pingback_enabled_;
920 scoped_ptr<net::URLFetcher> fetcher_;
921 scoped_refptr<SandboxedZipAnalyzer> analyzer_;
922 base::TimeTicks zip_analysis_start_time_;
923 bool finished_;
924 ClientDownloadRequest::DownloadType type_;
925 std::string client_download_request_data_;
926 base::CancelableTaskTracker request_tracker_; // For HistoryService lookup.
927 base::TimeTicks start_time_; // Used for stats.
928 base::TimeTicks timeout_start_time_;
929 base::TimeTicks request_start_time_;
930 base::WeakPtrFactory<CheckClientDownloadRequest> weakptr_factory_;
932 DISALLOW_COPY_AND_ASSIGN(CheckClientDownloadRequest);
935 DownloadProtectionService::DownloadProtectionService(
936 SafeBrowsingService* sb_service,
937 net::URLRequestContextGetter* request_context_getter)
938 : request_context_getter_(request_context_getter),
939 enabled_(false),
940 binary_feature_extractor_(new BinaryFeatureExtractor()),
941 download_request_timeout_ms_(kDownloadRequestTimeoutMs),
942 feedback_service_(new DownloadFeedbackService(
943 request_context_getter, BrowserThread::GetBlockingPool())) {
945 if (sb_service) {
946 ui_manager_ = sb_service->ui_manager();
947 database_manager_ = sb_service->database_manager();
951 DownloadProtectionService::~DownloadProtectionService() {
952 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
953 CancelPendingRequests();
956 void DownloadProtectionService::SetEnabled(bool enabled) {
957 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
958 if (enabled == enabled_) {
959 return;
961 enabled_ = enabled;
962 if (!enabled_) {
963 CancelPendingRequests();
967 void DownloadProtectionService::CheckClientDownload(
968 content::DownloadItem* item,
969 const CheckDownloadCallback& callback) {
970 scoped_refptr<CheckClientDownloadRequest> request(
971 new CheckClientDownloadRequest(item, callback, this,
972 database_manager_,
973 binary_feature_extractor_.get()));
974 download_requests_.insert(request);
975 request->Start();
978 void DownloadProtectionService::CheckDownloadUrl(
979 const content::DownloadItem& item,
980 const CheckDownloadCallback& callback) {
981 DCHECK(!item.GetUrlChain().empty());
982 scoped_refptr<DownloadUrlSBClient> client(
983 new DownloadUrlSBClient(item, callback, ui_manager_, database_manager_));
984 // The client will release itself once it is done.
985 BrowserThread::PostTask(
986 BrowserThread::IO,
987 FROM_HERE,
988 base::Bind(&DownloadUrlSBClient::StartCheck, client));
991 bool DownloadProtectionService::IsSupportedDownload(
992 const content::DownloadItem& item,
993 const base::FilePath& target_path) const {
994 // Currently, the UI is only enabled on Windows. On Mac we send the ping but
995 // ignore the result (see ifdef in FinishRequest). On Linux we still
996 // want to show the dangerous file type warning if the file is possibly
997 // dangerous which means we have to always return false here.
998 #if defined(OS_WIN)
999 DownloadCheckResultReason reason = REASON_MAX;
1000 ClientDownloadRequest::DownloadType type =
1001 ClientDownloadRequest::WIN_EXECUTABLE;
1002 return (CheckClientDownloadRequest::IsSupportedDownload(
1003 item, target_path, &reason, &type) &&
1004 (ClientDownloadRequest::CHROME_EXTENSION != type));
1005 #else
1006 return false;
1007 #endif
1010 DownloadProtectionService::ClientDownloadRequestSubscription
1011 DownloadProtectionService::RegisterClientDownloadRequestCallback(
1012 const ClientDownloadRequestCallback& callback) {
1013 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
1014 return client_download_request_callbacks_.Add(callback);
1017 void DownloadProtectionService::CancelPendingRequests() {
1018 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
1019 for (std::set<scoped_refptr<CheckClientDownloadRequest> >::iterator it =
1020 download_requests_.begin();
1021 it != download_requests_.end();) {
1022 // We need to advance the iterator before we cancel because canceling
1023 // the request will invalidate it when RequestFinished is called below.
1024 scoped_refptr<CheckClientDownloadRequest> tmp = *it++;
1025 tmp->Cancel();
1027 DCHECK(download_requests_.empty());
1030 void DownloadProtectionService::RequestFinished(
1031 CheckClientDownloadRequest* request) {
1032 DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
1033 std::set<scoped_refptr<CheckClientDownloadRequest> >::iterator it =
1034 download_requests_.find(request);
1035 DCHECK(it != download_requests_.end());
1036 download_requests_.erase(*it);
1039 void DownloadProtectionService::ShowDetailsForDownload(
1040 const content::DownloadItem& item,
1041 content::PageNavigator* navigator) {
1042 GURL learn_more_url(chrome::kDownloadScanningLearnMoreURL);
1043 learn_more_url = google_util::AppendGoogleLocaleParam(
1044 learn_more_url, g_browser_process->GetApplicationLocale());
1045 navigator->OpenURL(
1046 content::OpenURLParams(learn_more_url,
1047 content::Referrer(),
1048 NEW_FOREGROUND_TAB,
1049 ui::PAGE_TRANSITION_LINK,
1050 false));
1053 namespace {
1054 // Escapes a certificate attribute so that it can be used in a whitelist
1055 // entry. Currently, we only escape slashes, since they are used as a
1056 // separator between attributes.
1057 std::string EscapeCertAttribute(const std::string& attribute) {
1058 std::string escaped;
1059 for (size_t i = 0; i < attribute.size(); ++i) {
1060 if (attribute[i] == '%') {
1061 escaped.append("%25");
1062 } else if (attribute[i] == '/') {
1063 escaped.append("%2F");
1064 } else {
1065 escaped.push_back(attribute[i]);
1068 return escaped;
1070 } // namespace
1072 // static
1073 void DownloadProtectionService::GetCertificateWhitelistStrings(
1074 const net::X509Certificate& certificate,
1075 const net::X509Certificate& issuer,
1076 std::vector<std::string>* whitelist_strings) {
1077 // The whitelist paths are in the format:
1078 // cert/<ascii issuer fingerprint>[/CN=common_name][/O=org][/OU=unit]
1080 // Any of CN, O, or OU may be omitted from the whitelist entry, in which
1081 // case they match anything. However, the attributes that do appear will
1082 // always be in the order shown above. At least one attribute will always
1083 // be present.
1085 const net::CertPrincipal& subject = certificate.subject();
1086 std::vector<std::string> ou_tokens;
1087 for (size_t i = 0; i < subject.organization_unit_names.size(); ++i) {
1088 ou_tokens.push_back(
1089 "/OU=" + EscapeCertAttribute(subject.organization_unit_names[i]));
1092 std::vector<std::string> o_tokens;
1093 for (size_t i = 0; i < subject.organization_names.size(); ++i) {
1094 o_tokens.push_back(
1095 "/O=" + EscapeCertAttribute(subject.organization_names[i]));
1098 std::string cn_token;
1099 if (!subject.common_name.empty()) {
1100 cn_token = "/CN=" + EscapeCertAttribute(subject.common_name);
1103 std::set<std::string> paths_to_check;
1104 if (!cn_token.empty()) {
1105 paths_to_check.insert(cn_token);
1107 for (size_t i = 0; i < o_tokens.size(); ++i) {
1108 paths_to_check.insert(cn_token + o_tokens[i]);
1109 paths_to_check.insert(o_tokens[i]);
1110 for (size_t j = 0; j < ou_tokens.size(); ++j) {
1111 paths_to_check.insert(cn_token + o_tokens[i] + ou_tokens[j]);
1112 paths_to_check.insert(o_tokens[i] + ou_tokens[j]);
1115 for (size_t i = 0; i < ou_tokens.size(); ++i) {
1116 paths_to_check.insert(cn_token + ou_tokens[i]);
1117 paths_to_check.insert(ou_tokens[i]);
1120 std::string issuer_fp = base::HexEncode(issuer.fingerprint().data,
1121 sizeof(issuer.fingerprint().data));
1122 for (std::set<std::string>::iterator it = paths_to_check.begin();
1123 it != paths_to_check.end(); ++it) {
1124 whitelist_strings->push_back("cert/" + issuer_fp + *it);
1128 // static
1129 GURL DownloadProtectionService::GetDownloadRequestUrl() {
1130 GURL url(kDownloadRequestUrl);
1131 std::string api_key = google_apis::GetAPIKey();
1132 if (!api_key.empty())
1133 url = url.Resolve("?key=" + net::EscapeQueryParamValue(api_key, true));
1135 return url;
1138 } // namespace safe_browsing