Example - 1

In the following example, an authentication flag is read from a browser cookie, thus allowing for external control of user state data.

authenticated = true;Cookie c = cookies[i];if (c.getName().equals("authenticated") && Boolean.TRUE.equals(c.getValue())) {}Cookie[] cookies = request.getCookies();for (int i =0; i< cookies.length; i++) {}

Example - 2

The following code uses input from an HTTP request to create a file name. The programmer has not considered the possibility that an attacker could provide a file name such as "../../tomcat/conf/server.xml", which causes the application to delete one of its own configuration files (CWE-22).

String rName = request.getParameter("reportName");File rFile = new File("/usr/local/apfr/reports/" + rName);...rFile.delete();

Example - 3

The following code uses input from a configuration file to determine which file to open and echo back to the user. If the program runs with privileges and malicious users can change the configuration file, they can use the program to read any file on the system that ends with the extension .txt.

fis = new FileInputStream(cfg.getProperty("sub")+".txt");amt = fis.read(arr);out.println(arr);

Example - 4

This program is intended to execute a command that lists the contents of a restricted directory, then performs other actions. Assume that it runs with setuid privileges in order to bypass the permissions check by the operating system.

// /* Raise privileges to those needed for accessing DIR. */// 
#define DIR "/restricted/directory"char cmd[500];sprintf(cmd, "ls -l %480s", DIR);RaisePrivileges(...);system(cmd);DropPrivileges(...);...

This code may look harmless at first, since both the directory and the command are set to fixed values that the attacker can't control. The attacker can only see the contents for DIR, which is the intended program behavior. Finally, the programmer is also careful to limit the code that executes with raised privileges.

However, because the program does not modify the PATH environment variable, the following attack would work:

<xhtml_ul><xhtml_li>The user sets the PATH to reference a directory under the attacker's control, such as "/my/dir/".</xhtml_li><xhtml_li>The attacker creates a malicious program called "ls", and puts that program in /my/dir</xhtml_li><xhtml_li>The user executes the program.</xhtml_li><xhtml_li>When system() is executed, the shell consults the PATH to find the ls program</xhtml_li><xhtml_li>The program finds the attacker's malicious program, "/my/dir/ls". It doesn't find "/bin/ls" because PATH does not contain "/bin/".</xhtml_li><xhtml_li>The program executes the attacker's malicious program with the raised privileges.</xhtml_li></xhtml_ul>

Example - 5

The following code segment implements a basic server that uses the "ls" program to perform a directory listing of the directory that is listed in the "HOMEDIR" environment variable. The code intends to allow the user to specify an alternate "LANG" environment variable. This causes "ls" to customize its output based on a given language, which is an important capability when supporting internationalization.

$ENV{$1} = $2;

SendOutput($stream, "FILEINFO: $_");
open($fh, "/bin/ls -l $ENV{HOMEDIR}|");while (<$fh>) {}close($fh);chomp;if (/^ENV ([\w\_]+) (.*)/) {}elsif (/^QUIT/) { ... }elsif (/^LIST/) {}$ENV{"HOMEDIR"} = "/home/mydir/public/";my $stream = AcceptUntrustedInputStream();while (<$stream>) {}

The programmer takes care to call a specific "ls" program and sets the HOMEDIR to a fixed value. However, an attacker can use a command such as "ENV HOMEDIR /secret/directory" to specify an alternate directory, enabling a path traversal attack (CWE-22). At the same time, other attacks are enabled as well, such as OS command injection (CWE-78) by setting HOMEDIR to a value such as "/tmp; rm -rf /". In this case, the programmer never intends for HOMEDIR to be modified, so input validation for HOMEDIR is not the solution. A partial solution would be an allowlist that only allows the LANG variable to be specified in the ENV command. Alternately, assuming this is an authenticated user, the language could be stored in a local file so that no ENV command at all would be needed.

While this example may not appear realistic, this type of problem shows up in code fairly frequently. See CVE-1999-0073 in the observed examples for a real-world example with similar behaviors.