How Hash-Primarily based Protected Searching Works in Google Chrome


By Rohit Bhatia, Mollie Bates, Google Chrome Safety

There are numerous threats a person faces when searching the online. Customers could also be tricked into sharing delicate info like their passwords with a deceptive or faux web site, additionally referred to as phishing. They could even be led into putting in malicious software program on their machines, referred to as malware, which may accumulate private knowledge and in addition maintain it for ransom. Google Chrome, henceforth referred to as Chrome, permits its customers to guard themselves from such threats on the web. When Chrome customers browse the online with Protected Searching protections, Chrome makes use of the Protected Searching service from Google to determine and chase away numerous threats.

Protected Searching works in numerous methods relying on the person’s preferences. In the commonest case, Chrome makes use of the privacy-conscious Replace API (Software Programming Interface) from the Protected Searching service. This API was developed with person privateness in thoughts and ensures Google will get as little details about the person’s searching historical past as potential. If the person has opted-in to “Enhanced Safety” (lined in an earlier submit) or “Make Searches and Searching Higher“, Chrome shares restricted extra knowledge with Protected Searching solely to additional enhance person safety.

This submit describes how Chrome implements the Replace API, with acceptable tips that could the technical implementation and particulars in regards to the privacy-conscious elements of the Replace API. This must be helpful for customers to know how Protected Searching protects them, and for builders to flick thru and perceive the implementation. We are going to cowl the APIs used for Enhanced Safety customers in a future submit.

Threats on the Web

When a person navigates to a webpage on the web, their browser fetches objects hosted on the web. These objects embody the construction of the webpage (HTML), the styling (CSS), dynamic conduct within the browser (Javascript), pictures, downloads initiated by the navigation, and different webpages embedded in the principle webpage. These objects, additionally referred to as assets, have an internet tackle which is known as their URL (Uniform Useful resource Locator). Additional, URLs might redirect to different URLs when being loaded. Every of those URLs can probably host threats equivalent to phishing web sites, malware, undesirable downloads, malicious software program, unfair billing practices, and extra. Chrome with Protected Searching checks all URLs, redirects or included assets, to determine such threats and defend customers.

Protected Searching Lists

Protected Searching gives an inventory for every risk it protects customers towards on the web. A full catalog of lists which are utilized in Chrome might be discovered by visiting chrome://safe-browsing/#tab-db-manager on desktop platforms.

An inventory doesn’t comprise unsafe net addresses, additionally known as URLs, in entirety; it might be prohibitively costly to maintain all of them in a tool’s restricted reminiscence. As a substitute it maps a URL, which might be very lengthy, by way of a cryptographic hash operate (SHA-256), to a singular mounted dimension string. This distinct mounted dimension string, referred to as a hash, permits an inventory to be saved effectively in restricted reminiscence. The Replace API handles URLs solely within the type of hashes and can be referred to as hash-based API on this submit.

Additional, an inventory doesn’t retailer hashes in entirety both, as even that will be too reminiscence intensive. As a substitute, barring a case the place knowledge will not be shared with Google and the listing is small, it incorporates prefixes of the hashes. We check with the unique hash as a full hash, and a hash prefix as a partial hash.

An inventory is up to date following the Replace API’s request frequency part. Chrome additionally follows a back-off mode in case of an unsuccessful response. These updates occur roughly each half-hour, following the minimal wait period set by the server within the listing replace response.

For these all in favour of searching related supply code, right here’s the place to look:

Supply Code

  1. GetListInfos() incorporates all of the lists, together with their related risk sorts, the platforms they’re used on, and their file names on disk.
  2. HashPrefixMap exhibits how the lists are saved and maintained. They’re grouped by the scale of prefixes, and appended collectively to permit fast binary search based mostly lookups.

How is hash-based URL lookup completed

For example of a Protected Searching listing, as an example that we now have one for malware, containing partial hashes of URLs identified to host malware. These partial hashes are typically 4 bytes lengthy, however for illustrative functions, we present solely 2 bytes.

['036b', '1a02', 'bac8', 'bb90']

Each time Chrome must examine the status of a useful resource with the Replace API, for instance when navigating to a URL, it doesn’t share the uncooked URL (or any piece of it) with Protected Searching to carry out the lookup. As a substitute, Chrome makes use of full hashes of the URL (and a few mixtures) to lookup the partial hashes within the domestically maintained Protected Searching listing. Chrome sends solely these matched partial hashes to the Protected Searching service. This ensures that Chrome gives these protections whereas respecting the person’s privateness. This hash-based lookup occurs in three steps in Chrome:

Step 1: Generate URL Combos and Full Hashes

When Google blocks URLs that host probably unsafe assets by inserting them on a Protected Searching listing, the malicious actor can host the useful resource on a special URL. A malicious actor can cycle by way of numerous subdomains to generate new URLs. Protected Searching makes use of host suffixes to determine malicious domains that host malware of their subdomains. Equally, malicious actors may also cycle by way of numerous subpaths to generate new URLs. So Protected Searching additionally makes use of path prefixes to determine web sites that host malware at numerous subpaths. This prevents malicious actors from biking by way of subdomains or paths for brand new malicious URLs, permitting strong and environment friendly identification of threats.

To include these host suffixes and path prefixes, Chrome first computes the complete hashes of the URL and a few patterns derived from the URL. Following Protected Searching API’s URLs and Hashing specification, Chrome computes the complete hashes of URL mixtures by following these steps:

  1. First, Chrome converts the URL right into a canonical format, as outlined within the specification.
  2. Then, Chrome generates as much as 5 host suffixes/variants for the URL.
  3. Then, Chrome generates as much as 6 path prefixes/variants for the URL.
  4. Then, for the mixed 30 host suffixes and path prefixes mixtures, Chrome generates the complete hash for every mixture.

Supply Code

  1. V4LocalDatabaseManager::CheckBrowseURL is an instance which performs a hash-based lookup.
  2. V4ProtocolManagerUtil::UrlToFullHashes creates the varied URL mixtures for a URL, and computes their full hashes.


As an illustration, as an example {that a} person is attempting to go to The canonical url is The host suffixes to be tried are, and The trail prefixes are / and /blah. The 4 mixed URL mixtures are,,, and

url_combinations = ["", "","", ""]
full_hashes = ['1a02…28', 'bb90…9f', '7a9e…67', 'bac8…fa']

Step 2: Search Partial Hashes in Native Lists

Chrome then checks the complete hashes of the URL mixtures towards the domestically maintained Protected Searching lists. These lists, which comprise partial hashes, don’t present a decisive malicious verdict, however can shortly determine if the URL is taken into account not malicious. If the complete hash of the URL doesn’t match any of the partial hashes from the native lists, the URL is taken into account secure and Chrome proceeds to load it. This occurs for greater than 99% of the URLs checked.

Supply Code

  1. V4LocalDatabaseManager::GetPrefixMatches will get the matching partial hashes for the complete hashes of the URL and its mixtures.


Chrome finds that three full hashes 1a02…28, bb90…9f, and bac8…fa match native partial hashes. We notice that that is for demonstration functions, and a match right here is uncommon.

Step 3: Fetch Matching Full Hashes

Subsequent, Chrome sends solely the matching partial hash (not the complete URL or any explicit a part of the URL, and even their full hashes), to the Protected Searching service’s methodology. In response, it receives the complete hashes of all malicious URLs for which the complete hash begins with one of many partial hashes despatched by Chrome. Chrome checks the fetched full hashes with the generated full hashes of the URL mixtures. If any match is discovered, it identifies the URL with numerous threats and their severities inferred from the matched full hashes.

Supply Code

  1. V4GetHashProtocolManager::GetFullHashes performs the lookup for the complete hashes for the matched partial hashes.


Chrome sends the matched partial hashes 1a02, bb90, and bac8 to fetch the complete hashes. The server returns full hashes that match these partial hashes, 1a02…28, bb90…ce, and bac8…01. Chrome finds that one of many full hashes matches with the complete hash of the URL mixture being checked, and identifies the malicious URL as internet hosting malware.


Protected Searching protects Chrome customers from numerous malicious threats on the web. Whereas offering these protections, Chrome faces challenges equivalent to constraints in reminiscence capability, community bandwidth utilization, and a dynamic risk panorama. Chrome can be conscious of the customers’ privateness selections, and shares little knowledge with Google.

In a observe up submit, we’ll cowl the extra superior protections Chrome gives to its customers who’ve opted in to “Enhanced Safety”.


Please enter your comment!
Please enter your name here