Google Hacking: 10 Simple Security Searches That Work

Google has become the de facto standard in the search arena. It’s easy, quick and powerful. For those same reasons that the general user has gravitated to Google, so have the hackers. And as we all know, if the hackers do Google Hacking, the security professionals need to utilize it as well. And it doesn’t hurt to have Johnny Long (with help from Ed Skoudis) showing you the ropes. Enjoy this highly informative book. We did!

Google Hacking Chapter 7 – Ten Simple Security Searches That Work

Solutions in this Chapter:

• Site
• intitle:index.of
• error | warning
• login | logon
• username | userid | employee.ID | “your username is”
• password | passcode | “your password is”
• admin | administrator
• –ext:html –ext:htm –ext:shtml –ext:asp –ext:php
• inurl:temp | inurl:tmp | inurl:backup | inurl:bak
• intranet | help.desk
• List of Sites

Introduction

Although we see literally hundreds of Google searches throughout this book, sometimes it’s nice to know there’s a few searches that give good results just about every time. In the context of security work, we’ll take a look at 10 searches that work fairly well during a security assessment, especially when combined with the site operator, which secures the first position in our list. As you become more and more comfortable with Google, you’ll certainly add to this list, modifying a few searches and quite possibly deleting a few, but the searches here should serve as a very nice baseline for your own top 10 list. Without further ado, let’s dig into some queries.

site

The site operator is absolutely invaluable during the information-gathering phase of an assessment. Combined with a host or domain name, this query presents results that can be overwhelming, to say the least. However, the site operator is meant to be used as a base search, not necessarily as a standalone search. Sure, it’s possible (and not entirely discouraged) to scan through every single page of results from this query, but in most cases it’s just downright impractical.

Important information can be gained from a straight-up site search, however. First, remember that Google lists results in page-ranked order. In other words, the most popular pages float to the top of the results. This means you can get a quick idea about what the rest of the Internet thinks is most worthwhile about a site. The implications of this information are varied, but at a basic level you can at least get an idea of the public image or consensus about an online presence by looking at what floats to the top. Outside the specific site search itself, it can be helpful to read into the context of links originating from other sites. If a link’s text says something to the effect of “CompanyXYZ sucks!” there’s a good chance that some discontent is breeding somewhere about CompanyXYZ.

As we saw in Chapter 5, the site search can also be used to gather information about the servers and hosts that a target hosts. Using simple reduction techniques, we can quickly get an idea about a target’s online presence. Consider the simple example of site:washingtonpost.com –site:www.washingtonpost.com shown in Figure 7.1.

Figure 7.1 Site Reduction Reveals Domain Names

This query effectively locates pages on the washingtonpost.com domain other than www.washingtonpost.com. Just from a first pass, Figure 7.1 shows three other domains: yp.washingtonpost.com, eg.washingtonpost.com, and topics.washingtonpost.com. Although one result lists washingtonpost.com as a server name (without the www prefix), a DNS lookup quickly reveals that it points to the same IP as washingtonpost.com, as expected. Google might be perfectly suited for performing reconnaissance, but it’s always a good idea to validate your Google findings whenever possible.

intitle:index.of

intitle:index.of is the universal search for directory listings. In most cases, this search applies only to Apache-based servers, but due to the overwhelming number of Apache-derived Web servers on the Internet, there’s a good chance that the server you’re profiling will be Apache-based. Regardless, directory listings are chock-full of juicy details, as we saw in Chapter 3. Firing an intitle:index.of query against a target is fast and easy and could produce a killer payoff.

error | warning

As we’ve seen throughout this book, error messages can reveal a great deal of information about a target. Often overlooked, error messages can provide insight into the application or operating system software a target is running, the architecture of the network the target is on, information about users on the system, and much more. Not only are error messages informative, they are prolific. A query of intitle:error results in over 55 million results, as shown in Figure 7.2.

Figure 7.2 The Word Error Is Very Common in a Document Title

Unfortunately, some error messages don’t actually display the word error, as shown in the SQL located with a query of “access denied for user”“using password” shown in Figure 7.3.

Figure 7.3 Where Errors Hide, Warnings Lurk

This error page reveals usernames, filenames, path information, IP addresses, and line numbers, yet the worderror does not occur anywhere on the page. Nearly as prolific as error messages, warning messages can be generated from application programs. In some cases, however, the word warning is specifically written into the text of a page to alert the Web user that something important has happened or is about to happen. Regardless of how they are generated, pages containing these words may be of interest during an assessment, as long as you don’t mind teasing out the results a bit.

login | logon

As we’ll see in Chapter 8, a login portal is a “front door” to a Web site. Login portals can reveal the software and operating system of a target, and in many cases “self-help” documentation is linked from the main page of a login portal. These documents are designed to assist users who run into problems during the login process. Whether the user has forgotten his or her password or even user-name, these documents can provide clues that might help an attacker, or in our case a security tester, gain access to the site.

Many times, documentation linked from login portals lists e-mail addresses, phone numbers, or URLs of human assistants who can help a troubled user regain lost access. These assistants, or help desk operators, are perfect targets for a social engineering attack. Even the smallest security testing team should not be without a social engineering whiz who could talk an Eskimo out of his thermal boxer shorts. The vast majority of all security systems has one common weakest link:a human behind a keyboard. The words login and logon are widely used on the Internet, occurring on over 12 million pages, as shown in Figure 7.4.

Figure 7.4 login and logon Locate Login Portals

Notice that the very first result for this query shows the words login trouble in the text of the page. This link provides help to users who have forgotten their login credentials. It’s exactly these types of links that security testers might use to gain access to a system.

username | userid |
employee.ID | “your username is”

As we’ll see in Chapter 9, there are many different ways to obtain a username from a target system. Even though a username is the less important half of most authentication mechanisms, it should at least be marginally protected from outsiders. Figure 7.5 shows that even sites that reveal very little information in the face of a barrage of probing Google queries return many potentially interesting results to this query. To avoid implying anything negative about the target used in this example, some details of the figure have been edited.

Figure 7.5 Even “Tight-Lipped” Sites Provide Login Portals

The mere existence of the word username in a result is not indicative of a vulnerability, but results from this query provide a starting point for an attacker. Since there’s no good reason to remove derivations of the wordusername from a site you protect, why not rely on this common set of words to at least get a foothold during an assessment?

password | passcode | “your password is”

The word password is so common on the Internet, there are over 73 million results for this one-word query. Launching a query for derivations of this word makes little sense unless you actually combine that search with the site operator.

During an assessment, it’s very likely that results for this query combined with a site operator will include pages that provide help to users who have forgotten their passwords. In some cases, this query will locate pages that provide policy information about the creation of a password. This type of information can be used in an intelligent-guessing or even a brute-force campaign against a password field.

Despite how this query looks, it’s quite uncommon for this type of query to return actual passwords. Passwords do exist on the Web, but this query isn’t well suited for locating them. (We’ll look at queries to locate password in Chapter 9.) Like the login portal and username queries, this query can provide an informational foothold into a system. Although this query is somewhat useless without the site operator, Figure 7.6 shows that the first hit for this query is a “forgotten password” page—exactly the type of page that can be informative.

Figure 7.6 Even Without site, This Query Can Locate User Login Help Pages

admin | administrator

The word administrator is often used to describe the person in control of a network or system. There are so many references to the word on the Web that a query for admin | administrator weighs in at over 15 million results. This suggests that these words will likely be referenced on a site you’re charged with assessing. However, the value of these and other words in a query does not lie in the number of results but in the contextual relevance of the words. In this case, the word administrator is used in several common ways, each of which can provide relevance during an assessment. For example, the word administrator is referenced in many error messages as shown in Figure 7.7.

Figure 7.7 Admin Query Tweaked and Focused

The phrase Contact your system administrator is a fairly common phrase on the Web, as are several basic derivations. A query such as “please contact your * administrator” will return results that reference local, company, site, department, server, system, network, database, e-mail, and even tennis administrators. If a Web user is told to contact an administrator, odds are that there’s data of at least moderate importance to a security tester.

The word administrator can also be used to locate administrative login pages, or login portals. (We’ll take a closer look at login portal detection in Chapter 8.) A query for “administrative login” returns 150,000 results, many of which are administrative login pages. A security tester can profile Web servers using seemingly insignificant clues found on these types of login pages. Most login portals provide clues to an attacker about what software is in use on the server and act as a magnet, drawing attackers who are armed with an exploit for that particular type of software. Remember that Google performs autostemming; a search for “admin login”returns approximately 1.3 million results, including results that were autostemmed to include the phraseadministrator login. As shown in Figure 7.8, many of the results are for administrative login pages.

Figure 7.8 admin login Reveals Administrative Login Pages

Another interesting use of the administrator derivations is to search for them in the URL of a page using an inurlsearch. If the word admin is found in the hostname, a directory name, or a filename within a URL, there’s a decent chance that the URL has some administrative function, making it interesting from a security standpoint.

-ext:html –ext:htm –ext:shtml –ext:asp –ext:php

The –ext:html –ext:htm –ext:shtml –ext:asp –ext:php query uses ext, a synonym for the filetype operator, and is a negative query. It returns no results when used alone and should be combined with a site operator to work properly. The idea behind this query is to exclude some of the most common Internet file types in an attempt to find files that might be more interesting for our purposes.

As you’ll see through this book, there are certainly lots of HTML, PHP, and ASP pages that reveal interesting information, but this chapter is about cutting to the chase, and that’s what this query attempts to do. The documents returned by this search often have great potential for document grinding, which we’ll explore in more detail in Chapter 10.The file extensions used in this search were selected very carefully. First, www.filext.com(one of the Internet’s best resources for all known file extensions) was consulted to obtain a list of every known file extension. Each entry in the list of over 8000 file extensions was converted into a Google query using thefiletype operator. For example, if we wanted to search for the PDF extension, we might use a query likefiletype:PDF PDF to get the number of known results on the Internet.This type of Google query was performed for each and every known file extension from filext.com, which can take quite some time, considering that the Google API key only allows 1000 searches per day. Once the results were gathered, they were sorted in descending order by the number of hits.The top 20 results of this query are shown in Table 7.1.

Table 7.1 Top 20 File Extensions on the Internet

File Extension	Approximate Number of Hits
HTML	17,800,000
PHP	16,500,000
HTM	16,100,000
ASP	15,400,000
PDF	11,600,000
CGI	11,100,000
CFM	9,870,000
SHTML	8,770,000
JSP	7,370,000
ASPX	7,110,000
PL	5,660,000
PHP3	3,870,000
DLL	3,340,000
SWF	2,260,000
PHTML	2,250,000
DOC	2,120,000
FCGI	1,850,000
TXT	1,700,000
MV	1,060,000
JHTML	990,000

This table reveals the most common file types on the Internet, according to Google. In an attempt to get to the juiciest documents fast, our query opts to ignore the most common server-generated pages, which end in HTML, PHP, HTM,ASP, and SHTML. Typically a query like this, submitted with a site operator, will reveal a list of results worth investigating. In some cases, this query will need to be refined, especially if the site uses a less common server-generated file extension. For example, consider this query combined with a site operator, as shown in Figure 7.9. (To protect the identity of the target, certain portions of the figure have been edited.)

Figure 7.9 A Base Search Combined with the site Operator

As revealed in the search results, this site uses the ASPX extension for some Web content. By adding –ext:aspxto the query and resubmitting it, that type of content is removed from the search results. This modified search reveals some interesting information, as shown in Figure 7.10.

Figure 7.10 New and Improved, Juicier and Tastier

By adding a common file extension used on this site, after a few pages of mediocre results we discover a page full of interesting information. Result line 1 reveals that the site supports the HTTPS protocol, a secured version of HTTP used to protect sensitive information. The mere existence of the HTTPS protocol often indicates that this server houses something worth protecting. Result line 1 also reveals several nested subdirectories (/research/files/summaries) that could be explored or traversed to located other information. This same line also reveals the existence of a PDF document dated the first quarter of 2003.

Result line 2 reveals the existence of what is most likely a development server named DEV. This server also contains subdirectories (/events/archives/strategiesNAM2003) that could be traversed to uncover more information. One of the subdirectory names, strategiesNAM2003, contains a the string 2003, most likely a reference to the year 2003. Using the incremental substitution technique discussed in Chapter 3, it’s possible to modify the year in this directory name to uncover similarly named directories. Result line 2 also reveals the existence of an attendee list that could be used to discover usernames, e-mail addresses, and so on.

Result line 3 reveals another machine name, JOBS, which contains a ColdFusion application that accepts parameters. Depending on the nature and security of this application, an attack based on user input might be possible. Result line 4 reveals new directory names, /help/emp, which could be traversed or fed into other third-party assessment applications.

The results continue, but the point is that once common, purposefully placed files are removed from a search, interesting information tends to float to the top. This type of reduction can save an attacker or a security technician a good deal of time in assessing a target.

inurl:temp | inurl:tmp | inurl:backup | inurl:bak

The inurl:temp | inurl:tmp | inurl:backup | inurl:bak query, combined with the site operator, searches for temporary or backup files or directories on a server. Although there are many possible naming conventions for temporary or backup files, this search focuses on the most common terms. Since this search uses the inurloperator, it will also locate files that contain these terms as file extensions, such as index.html.bak, for example. Modifying this search to focus on file extensions is tricky because this requires OR’ing the filetype operator (which is often flaky, since filetype also requires a search term that gets lost in the mess of ORs) and also limits our search, leaving out temporary or backup directories.

intranet | help.desk

The term intranet, despite more specific technical meanings, has become a generic term that describes a network confined to a small group. In most cases the term intranet describes a closed or private network, unavailable to the general public. However, many sites have configured portals that allow access to an intranet from the Internet, bringing this typically closed network one step closer to potential attackers.

In rare cases, private intranets have been discovered on the public Internet due to a network device misconfiguration. In these cases, network administrators were completely unaware that their private networks were accessible to anyone via the Internet. Most often, an Internet-connected intranet is only partially accessible from the outside. In these cases, filters are employed that only allow access to certain pages from specific addresses, presumably inside a facility or campus. There are two major problems with this type of configuration. First, it’s an administrative nightmare to keep track of the access rights of specific pages. Second, this is not true access control. This type of restriction can be bypassed very easily if an attacker gains access to a local proxy server, bounces a request off a local misconfigured Web server, or simply compromises a machine on the same network as trusted intranet users. Unfortunately, it’s nearly impossible to provide a responsible example of this technique in action. Each example we considered for this section was too easy for an attacker to reconstruct with a few simple Google queries.

Help desks have a bad reputation of being, well, too helpful. Since the inception of help desks, hackers have been donning alternate personalities in an attempt to gain sensitive information from unsuspecting technicians. Recently, help desk procedures have started to address the hacker threat by insisting that technicians validate callers before attempting to assist them. Most help desk workers will (or should) ask for identifying information such as usernames, Social Security numbers, employee numbers, and even PIN numbers to properly validate callers’ identities. Some procedures are better than others, but for the most part, today’s help desk technicians are at least aware of the potential threat that is posed by an imposter.

In Chapter 4, we discussed ways Google can be used to harvest the identification information a help desk may require, but the intranet | help.desk query is designed not to bypass help desk procedures but rather to locate pages describing help desk procedures. When this query is combined with a site search, the results could indicate the location of a help desk (Web page, telephone number, or the like), the information that might be requested by help desk technicians (which an attacker could gather before calling), and in many cases links that describe troubleshooting procedures. Self-help documentation is often rather verbose, and a crafty attacker can use the information in these documents to profile a target network or server. There are exceptions to every rule, but odds are that this query, combined with the site operator, will dig up information about a target that can feed a future attack.

Summary

There’s no such thing as the perfect list, but these 10 searches should serve you well as you seek to compile your own list of killer searches. It’s important to realize that a search that works against one target might not work well against other targets. Keep track of the searches that work for you, and try to reach some common ground about what works and what doesn’t. Automated tools, discussed in Chapters 11 and 12, can be used to feed longer lists of Google queries such as those found in the Google Hacking Database, but in some cases, simpler might be better. If you’re having trouble finding common ground in some queries that work for you, don’t hesitate to keep them in a list for use in one of the automated tools we’ll discuss later.

Solutions Fast Track

site

• The site operator is great for trolling through all the content Google has gathered for a target.
• This operator is used in conjunction with many of the other queries presented here to narrow the focus of the search to one target.

intitle:index.of

• The universal search for Apache-style directory listings.
• Directory listings provide a wealth of information for an attacker.

error | warning

• Error messages are also very revealing in just about every context.
• In some cases, warning text can provide important insight into the behind-the-scenes code used by a target.

login | logon

• This query locates login portals fairly effectively.
• It can also be used to harvest usernames and troubleshooting
  procedures.

username | userid | employee.ID | “your username is”

• This is one of the most generic searches for username harvesting.
• In cases where this query does not reveal usernames, the context around these words can reveal procedural information an attacker can use in later offensive action.

password | passcode | “your password is”

• This query reflects common uses of the word password.
• This query can reveal documents describing login procedures, password change procedures, and clues about password policies in use on the target.

admin | administrator

• Using the two most common terms for the owner or maintainer of a site, this query can also be used to reveal procedural information (“contact your administrator”) and even admin login portals.

ext:html –ext:htm –ext:shtml –ext:asp –ext:php

• This query, when combined with the site operator, gets the most common files out of the way to reveal more interesting documents.
• This query should be modified to reduce other common file types on a target-by-target basis.

inurl:temp | inurl:tmp | inurl:backup | inurl:bak

• This query locates backup or temporary files and directories.

intranet | help.desk

• This query locates intranet sites (which are often supposed to be protected from the general public) and help desk contact information and procedures.

Frequently Asked Questions

The following Frequently Asked Questions, answered by the authors of this book, are designed to both measure your understanding of the concepts presented in this chapter and to assist you with real-life implementation of these concepts. To have your questions about this chapter answered by the author, browse towww.syngress.com/solutions and click on the “Ask the Author” form. You will also gain access to thousands of other FAQs at ITFAQnet.com.

Q: If automation is an option, what’s so great about 10 measly searches?

A: Automation tools, such as those discussed in Chapters 11 and 12, have their place. However, the vast majority of the searches covered in large query lists are very specific searches that target a very small minority of Internet sites. Although the effects of these specific queries are often devastating, it’s often nice to have a short list of powerful searches to get the creative juices flowing during an assessment, especially if you’ve reached a dead end using more conventional means.

Q: Doesn’t it make more sense to base a list like this off a more popular list like the SANS Top 20 list atwww.sans.org/top20?

A: There’s nothing wrong with the SANS Top 20 list, except for the fact that the vast majority of the items on the list describe vulnerabilities that are not Web-based. This means that in most cases the vulnerabilities described there cannot be detected or exploited via Web-based services such as Google.