There are 3 vulnerabilities in poppler and 1 in Xpdf that need CVE
attention. Can you please provide CVE's for the following?
- Race condition on temporary file (Windows) / Insecure temporary file
(other non-Unix OS's), affecting poppler and Xpdf (reported by Pedro
Ribeiro, unfixed in poppler, unfixed in Xpdf) -> Not sure if this is
one or two vulnerabilities?
- Stack based buffer overflow, affecting poppler in the utils section
(reported by Daniel Kahn Gillmor, fixed in poppler 0.24.2)
- User controlled format string, affecting poppler in the utils
section (reported by Daniel Kahn Gillmor and Pedro Ribeiro, fixed in
poppler 0.24.3)
Note that the poppler maintainers are aware of the unfixed issue. Xpdf
upstream appears to be dead since 2011 so I have not attempted to
contact them.
Details on the vulnerabilities are below.
The first vulnerability is use of insecure temporary file for non-Unix
OS's. As per the code comments, the maintainers are aware of this and
welcome patches from anyone who knows of a better way to create temp
files in Windows / other OS's.
I have also checked Xpdf and the same vulnerable code is present, so
the bug must be pretty old and all releases of poppler since forking
from Xpdf should be affected.
======================================================================
Vulnerability: Race condition on temporary file access / Insecure
Temporary File (CWE-363 / CWE-377)
Filename(line): poppler-0.24.2/goo/gfile.cc(340-395)
Code snippet:
There is a race condition and use of a insecure temporary file in the
openTempFile function that enables an attacker to replace the temporary file with
a symlink of his/her choosing. This only happens on non-Unix OS's (old MacOS, Windows, etc).
GBool openTempFile(GooString **name, FILE **f, const char *mode) {
#if defined(_WIN32)
//---------- Win32 ----------
char *tempDir;
GooString *s, *s2;
FILE *f2;
int t, i;
// this has the standard race condition problem, but I haven't found
// a better way to generate temp file names with extensions on
// Windows
if ((tempDir = getenv("TEMP"))) {
s = new GooString(tempDir);
s->append('\\');
} else {
s = new GooString();
}
s->appendf("x_{0:d}_{1:d}_",
(int)GetCurrentProcessId(), (int)GetCurrentThreadId());
t = (int)time(NULL);
for (i = 0; i < 1000; ++i) {
s2 = s->copy()->appendf("{0:d}", t + i);
if (!(f2 = fopen(s2->getCString(), "r"))) {
if (!(f2 = fopen(s2->getCString(), mode))) {
delete s2;
delete s;
return gFalse;
}
*name = s2;
*f = f2;
delete s;
return gTrue;
}
fclose(f2);
delete s2;
}
delete s;
return gFalse;
#elif defined(VMS) || defined(__EMX__) || defined(ACORN) || defined(MACOS)
//---------- non-Unix ----------
char *s;
// There is a security hole here: an attacker can create a symlink
// with this file name after the tmpnam call and before the fopen
// call. I will happily accept fixes to this function for non-Unix
// OSs.
if (!(s = tmpnam(NULL))) {
return gFalse;
}
*name = new GooString(s);
if (!(*f = fopen((*name)->getCString(), mode))) {
delete (*name);
*name = NULL;
return gFalse;
}
return gTrue;
======================================================================
The second vulnerability is a buffer overflow in the pdfseparate
utility, and was reported by Daniel Kahn Gillmor. The buffer overflow
was fixed in poppler 0.24.2 as per commit in [1].
The third vulnerability user controlled format string, which was
reported by Daniel Kahn Gillmor and Pedro Ribeiro separately to the
poppler maintainers. This vulnerability was fixed on poppler 0.24.3 as
per the commit in [2].
More details on the format string are below:
======================================================================
Vulnerability: Uncontrolled format string (CWE-124)
Filename(line): poppler-0.24.2/utils/pdfseparate.cc(70)
Code snippet:
bool extractPages (const char *srcFileName, const char *destFileName) {
char pathName[4096];
GooString *gfileName = new GooString (srcFileName);
PDFDoc *doc = new PDFDoc (gfileName, NULL, NULL, NULL);
...
if (firstPage != lastPage && strstr(destFileName, "%d") == NULL) {
error(errSyntaxError, -1, "'{0:s}' must contain '%%d' if more than
one page should be extracted", destFileName);
return false;
}
for (int pageNo = firstPage; pageNo <= lastPage; pageNo++) {
snprintf (pathName, sizeof (pathName) - 1, destFileName, pageNo);
^ function parameter passed as format string
The function is called by main in line 110 directly passing the arguments:
ok = extractPages (argv[1], argv[2]);
^ destFileName parameter
PoC:
./pdfseparate -f 1 -l 1 aPdfFile.pdf "%x%x%x%x%x%x%n"
======================================================================
Regards,
Pedro
