Vulnerability analysis, sometimes called vulnerability scanning, is the act of determining which security holes and vulnerabilities may be applicable to the target network. In order to do this, we examine identified machines within the target network to identify all open ports and the operating systems and applications the hosts are running (including version number, patch level, and service pack). In addition, we compare this information with several Internet vulnerability databases to ascertain what current vulnerabilities and exploits may be applicable to the target network.

Given the time constraint we may be under during an engagement and the number of hosts within scope, it may be necessary to focus initially on critical hosts. However, if any paring down of the target list needs to be done, it is usually done during the next step.

Note: It is important to take into consideration that the ping results do not authoritatively show that a host is down. In light of this, if there is any doubt as to whether the target(s) are effectively filtered or protected from ping or are actually down, we recommend continuing with a port scan. Keep the number of ports in such scans down as these scans tend to take a longer amount of time. If it is necessary to scan a large number of ports on unresponsive hosts, it is best to do this overnight.

At the end of this stage, we like to be able to document all target hosts (alive and otherwise) in a table along with the OS, IP address, running applications, any banner information available, and known vulnerabilities. This information is useful both during the exploitation stage and for presentation to the client so that the client becomes aware of the vulnerabilities on the network and the amount of information an outsider can gather prior to compromising the network.

OS Identification

By identifying the operating system, we can attempt to predict services that may be running on the host and tailor our port scans based on this information. Nmap, the leading tool used to perform OS identification, does this by analyzing the response of the target's TCP stack to the packets Nmap sends out. Various RFCs govern how the TCP stack should respond when queried. However, implementation details are left to the vendor. Therefore, differences in how vendors satisfy the RFCs allow them to be identified. While this method is not foolproof, Nmap's OS detection is fairly reliable and well accepted by industry. Changing a computer's OS signature is possible but not trivial, and it has not been our experience that companies perform this level of masking.

OS identification goes a long way in performing network enumeration and vulnerability scanning. As soon as we know the OS of a particular machine, we can begin to generate a list of potential holes and vulnerabilities—often from the vendor's own Web site. For example, as soon as we know a machine is Windows NT, we can check whether TCP port 139 is open and attempt a null connection to the IPC$ share. If we identify a UNIX box, we can look for the X Windows ports (6000–6063).

Port Scanning

Port scans attempt to determine whether a listening service is listening on a given port. There are many variations in performing a port scan. We describe those that have been the most useful for us. Depending on the level of stealth you want to maintain, there are a variety of scan types you can use, TCP SYN scan is the most popular of the stealthy port scans.

A scan of all possible ports (1–65535) is the most comprehensive and offers the most information on the target, but it also is the most time consuming and maximizes the chances of being identified by the target. Such a port scan is usually performed only by beginner hackers. If you do elect to scan all ports, we strongly recommend that the scan be performed over several sessions, each with a smaller port range. Often we perform a comprehensive scan at the end of the engagement when stealth is no longer necessary. This helps identify any services that we may have missed during the surgical port scans.

If you are not attempting to avoid detection and are just trying to identify weaknesses in the target system (for example, if your client's security staff is fully aware of your penetration testing efforts), then there is no problem with scanning all ports at once. However, it will take a long time to return information. The test becomes more efficient if you can review results while simultaneously scanning new systems.

Luckily, there are several alternatives to a full port scan. We can stick to the basic known ports, 1–1024, and add a few other ports we know to be relevant to the client, such as the X Windows ports (6000–6063) on a UNIX machine. Reviewing the /etc/services file on a UNIX machine also provides a good listing of ports to scan. We can also create a list of ports that support applications with known vulnerabilities we may wish to exploit, such as FTP, telnet, and RealSecure (ports 21, 23, and 2998, respectively). Most scanners give you the ability to scan both TCP and UDP ports. Often, UDP ports are simply ignored since they are less common, but they can be just as vulnerable. Since UDP is a connectionless protocol, UDP port scans are generally considered less reliable.

Ultimately, you should develop a port list that you are comfortable using on any given network and modify the list to fit the particular network you are going to scan. Specifically, create a generic list, and remove NT-specific ports for a UNIX network or vice versa. Nmap is distributed with a port list of several known ports and can serve as a starting point for such a generic list. To this list, add ports each time you find another one that is associated with an application featuring a vulnerability you are aware of or a security hole you can exploit. Remove ports that are not related to weaknesses, vulnerabilities, or information gathering. While maintaining such a list demands continuous testing, the more port scanning you do, the more relevant the information your own list will return.

As previously mentioned, on UNIX environments, Nmap is the tool of our choice for port scanning and is considered to be the premier port scanner available (as well as a reliable OS detection tool). SuperScan and 7th Sphere are effective port scanners for the NT environment but do not include OS detection. (As mentioned, an Nmap for the Windows environment is under development.)

Port scanners generally ping hosts before scanning and only scan those hosts that respond to pings. If there is any doubt as to the validity of the ping results, then set the port scanners to scan hosts unresponsive to ping. As a consequence, the port scanning will take more time.

The legality of port scanning has long been a topic of discussion in the security community. Some professionals have indicated port scanning is no more than driving down a street looking at houses and noticing which windows are open. However, port scanning without permission is clearly unethical and will always be alarming as a possible prelude to an attack.

Application Enumeration

From the results of port scanning, we gain a list of open ports on the target machines. An open port does not entirely indicate what listening service may be active. Ports below 1024 have been assigned to various services and if these are found open, they generally indicate the assigned service. Additionally, other applications have been run on certain ports for so long that they have become the de facto standard, such as port 65301 for pcAnywhere and 26000 for Quake. Of course system administrators can change the port a service runs on in an attempt to “hide it” (an example of security through obscurity). Therefore, we attempt to connect to the open port and grab a banner to verify the service running.

Knowing which applications the target hosts are running goes a long way toward performing vulnerability analysis. Just as with knowing the OS, we can run the list of applications through the Internet and find a list of known vulnerabilities and exploits for these applications—again, often from the vendors themselves.

Application enumeration also involves banner grabbing. Several applications (including telnet, FTP, HTTP, SNMP, and a host of others) identify themselves and their version in their user name/password challenge screen. This information is called a banner and is very helpful in identifying running applications. We generally record any banners we come across during our penetration testing. This can be done with many applications, including netcat, which runs on either the UNIX or Windows command line; telnet; and What's Running, a Windows GUI tool.

A benefit of What's Running is that the banner is placed in a window from which it can be copied into a file or edited.

5.2.4 Internet Research

Once the list of applications is known, the next step is to research the list and determine what vulnerabilities exist. As you perform penetration tests, you become familiar with certain popular vulnerabilities and can quickly determine whether an application is vulnerable. However, it is important to keep in mind that new vulnerabilities are posted on a daily basis, and you should check your favorite vulnerability databases for all the applications, services, and operating systems you find on each engagement.

The popular Internet sites with vulnerability databases are the Bugtraq lists, Packetstorm (www.packetstormsecurity.org), and SecurityFocus (www.securityfocus.com). You will become comfortable with these and other sites that offer the ability to quickly search for vulnerabilities. Even though there is some amount of duplication among them, you may want to check multiple sites since no one repository is entirely comprehensive. Each site has its own network of users and supporters who update the databases.

Once we have identified vulnerabilities and documented them in our table, we download the exploit code (if applicable) to use in the next phase of the penetration test.