CERT C: Rec. STR02-C

Sanitize data passed to complex subsystems

Description

Rule Definition

Sanitize data passed to complex subsystems.¹

Polyspace Implementation

The rule checker checks for these issues:

• Execution of externally controlled command.

• Command executed from externally controlled path.

• Library loaded from externally controlled path.

Extend Checker

A default Bug Finder analysis might not flag this checker for certain inputs that originate outside of the current analysis boundary. See Sources of Tainting in a Polyspace Analysis. To consider any data that does not originate in the current scope of Polyspace analysis as tainted, use the command line option -consider-analysis-perimeter-as-trust-boundary.

Examples

Execution of externally controlled command

Issue

Execution of externally controlled command checks for commands that are fully or partially constructed from externally controlled input.

Risk

Attackers can use the externally controlled input as operating system commands, or arguments to the application. An attacker could read or modify sensitive data can be read or modified, execute unintended code, or gain access to other aspects of the program.

Fix

Validate the inputs to allow only intended input values. For example, create a list of acceptable inputs and compare the input against this list.

Example - Call External Command

#include "stdlib.h"
#include "stdio.h"
#include "string.h"
#include "unistd.h"
#include "dlfcn.h"
#include "limits.h"
#define MAX 128


void taintedexternalcmd(void)
{
	char* usercmd;
	fgets(usercmd,MAX,stdin);
	char cmd[MAX] = "/usr/bin/cat ";
	strcat(cmd, usercmd);
	system(cmd);  //Noncompliant
}

This example function calls a command from a user input without checking the command variable.

Correction — Use a Predefined Command

One possible correction is to use a switch statement to run a predefined command, using the user input as the switch variable.

#define _XOPEN_SOURCE
#define _GNU_SOURCE

#include "stdlib.h"
#include "stdio.h"
#include "string.h"
#include "unistd.h"
#include "dlfcn.h"
#include "limits.h"

enum {
    SIZE10  =  10,
    SIZE100 = 100,
    SIZE128 = 128
};
enum { CMD0 = 1, CMD1, CMD2 };

void taintedexternalcmd(void)
{
    int usercmd = strtol(getenv("cmd"),NULL,10);
    char cmd[SIZE128] = "/usr/bin/cat ";

    switch(usercmd) {
        case CMD0:
            strcat(cmd, "*.c");
            break;
        case CMD1:
            strcat(cmd, "*.h");
            break;
        case CMD2:
            strcat(cmd, "*.cpp");
            break;
        default:
            strcat(cmd, "*.c");
    }
    system(cmd); 
}

Command executed from externally controlled path

Issue

Command executed from externally controlled path checks the path of commands that the application controls. If the path of a command is from or constructed from external sources, Bug Finder flags the command function.

Risk

An attacker can:

• Change the command that the program executes, possibly to a command that only the attack can control.

• Change the environment in which the command executes, by which the attacker controls what the command means and does.

Fix

Before calling the command, validate the path to make sure that it is the intended location.

Example - Executing Path from Environment Variable

#include <stdlib.h>
#include <stdio.h>
#include <string.h>

enum {
    SIZE10  =  10,
    SIZE100 = 100,
    SIZE128 = 128
};

void bug_taintedpathcmd() {
    char cmd[SIZE128] = "";
    char* userpath = getenv("MYAPP_PATH");

    strncpy(cmd, userpath, SIZE100);
    strcat(cmd, "/ls *");
    /* Launching command */
    system(cmd);  //Noncompliant
}

This example obtains a path from an environment variable MYAPP_PATH. system runs a command from that path without checking the value of the path. If the path is not the intended path, your program executes in the wrong location.

Correction — Use Trusted Path

One possible correction is to use a list of allowed paths to match against the environment variable path.

#include <stdlib.h>
#include <stdio.h>
#include <string.h>

enum {
    SIZE10  =  10,
    SIZE100 = 100,
    SIZE128 = 128
};

/* Function to sanitize a string */
int sanitize_str(char* s, size_t n) {
    int res = 0;
    /* String is ok if */
    if (s && n>0 && n<SIZE128) {
        /* - string is not null                     */
        /* - string has a positive and limited size */
        s[n-1] = '\0';  /* Add a security \0 char at end of string */
        /* Tainted pointer detected above, used as "firewall" */
        res = 1;
    }
    return res;
}

/* Authorized path ids */
enum { PATH0=1, PATH1, PATH2 };

void taintedpathcmd() {
    char cmd[SIZE128] = "";

    char* userpathid = getenv("MYAPP_PATH_ID");
    if (sanitize_str(userpathid, SIZE100)) {
        int pathid = atoi(userpathid);


        char path[SIZE128] = "";
        switch(pathid) {
            case PATH0:
                strcpy(path, "/usr/local/my_app0");
                break;
            case PATH1:
                strcpy(path, "/usr/local/my_app1");
                break;
            case PATH2:
                strcpy(path, "/usr/local/my_app2");
                break;
            default:
                /* do nothing */
		break;
        }
        if (strlen(path)>0) {
            strncpy(cmd, path, SIZE100);
            strcat(cmd, "/ls *");
            system(cmd);    
        }
    }
}

Library loaded from externally controlled path

Issue

Library loaded from externally controlled path looks for libraries loaded from fixed or controlled paths. If unintended actors can control one or more locations on this fixed path, Bug Finder raises a defect.

Risk

If an attacker knows or controls the path that you use to load a library, the attacker can change:

• The library that the program loads, replacing the intended library and commands.

• The environment in which the library executes, giving unintended permissions and capabilities to the attacker.

Fix

When possible, use hard-coded or fully qualified path names to load libraries. It is possible the hard-coded paths do not work on other systems. Use a centralized location for hard-coded paths, so that you can easily modify the path within the source code.

Another solution is to use functions that require explicit paths. For example, system() does not require a full path because it can use the PATH environment variable. However, execl() and execv() do require the full path.

Example - Call Custom Library

#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <dlfcn.h>
#include <limits.h>

enum {
    SIZE10  =  10,
    SIZE100 = 100,
    SIZE128 = 128
};

void* taintedpathlib() {
    void* libhandle = NULL;
    char lib[SIZE128] = "";
    char* userpath = getenv("LD_LIBRARY_PATH");
    strncpy(lib, userpath, SIZE128);
    strcat(lib, "/libX.so");
    libhandle = dlopen(lib, 0x00001); //Noncompliant
    return libhandle;
}

This example loads the library libX.so from an environment variable LD_LIBRARY_PATH. An attacker can change the library path in this environment variable. The actual library you load could be a different library from the one that you intend.

Correction — Change and Check Path

One possible correction is to change how you get the library path and check the path of the library before opening the library. This example receives the path as an input argument. Then the path is checked to make sure the library is not under /usr/.

#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <dlfcn.h>
#include <limits.h>

enum {
    SIZE10  =  10,
    SIZE100 = 100,
    SIZE128 = 128
};

/* Function to sanitize a string */
int sanitize_str(char* s, size_t n) {
    /* strlen is used here as a kind of firewall for tainted string errors */
    int res = (strlen(s) > 0 && strlen(s) < n);
    return res;
}
void* taintedpathlib(char* userpath) {
    void* libhandle = NULL;
    if (sanitize_str(userpath, SIZE128)) {
        char lib[SIZE128] = "";

        if (strncmp(userpath, "/usr", 4)!=0) {
            strncpy(lib, userpath, SIZE128);
            strcat(lib, "/libX.so");
            libhandle = dlopen(lib, RTLD_LAZY);    
        }
    }
    return libhandle;
}

Check Information

Group: Rec. 07. Characters and Strings (STR)

Version History

Introduced in R2019a