Buffer Overflow (BOF) Examples

CVE-2014-0160 (Heartbleed)

BF Taxonomy

Cause: Input not checked properly leads to Data Exceeds Array (specifically Too Much Data)

Attributes:
Access: Read
Boundary: Above
Location: Heap
Data Size: Huge
Excursion: Continuous

Consequence: IEX (if not had been cleared - CWE-226)

BF Description

"Input not checked properly leads to too much data, where huge data is read from the heap in a continuous excursion above the array boundary, which may be exploited for IEX (if not had been cleared)." [2]

CVE Description

"The (1) TLS and (2) DTLS implementations in OpenSSL 1.0.1 before 1.0.1g do not properly handle Heartbeat Extension packets, which allows remote attackers to obtain sensitive information from process memory via crafted packets that trigger a buffer over-read, as demonstrated by reading private keys, related to d1_both.c and t1_lib.c, aka the Heartbleed bug." [1]

Analysis

The following analysis is based on information in [1,3,4,5]. A user has to send to the software data, and a number called payload that is the length of that data. The software has to send the data received back to the user. In the Heartbleed attack, a malicious user gives payload a value that can be as large as 65535+1+2+16, and sends data having a number of bytes that is much less than payload, and can be as small as 1. The software stores that data in an array that it allocated for that purpose. The size of that array is much less than 65535+1+2+16. The software does not check the data and the value of payload in order to make sure that the number of bytes of data is equal to payload. The software therefore assumes that those numbers are equal. The software reads, using memcpy, payload consecutive bytes from that array, beginning at its first byte, (continuous excursion) and sends them to the malicious user. This results in reading a large number of bytes beyond the end of the allocated array. The software did not clear the memory that it read from beyond the allocated array. Therefore, the data read and sent to the malicious user by the software includes sensitive information.

Source Code

Code With Bug

1 2 3 4 5 6 7 8 hbtype = *p++; n2s(p, payload); pl = p;

Code With Fix

1 2 3 4 5 6 7 8 /* Read type and payload length first */ if (1 +2 + 16 > s->s3->rrec.length)  return 0; /* silently discard */ hbtype = *p++; n2s(p, payload); if (1 + 2 + payload + 16 > s->s3->rrec.length)  return 0; /* silently discard per RFC 6520 sec. 4 */ pl = p;

CVE-2015-0235 (Ghost)

BF Taxonomy

Cause: Incorrect Calculation (Missing Factor) leads to Array Too Small

Attributes:
Access: Write
Boundary: Above
Location: Heap
Data Size: Some
Excursion: Continuous

Consequence: Arbitrary Code Execution leading to DoS

BF Description

"Incorrect calculation (missing factor) leads to array too small, where some data is written to the heap in a continuous excursion above the array boundary, which may be exploited for arbitrary code execution, leading to DoS." [2]

CVE Description

"Heap-based buffer overflow in the __nss_hostname_digits_dots function in glibc 2.2, and other 2.x versions before 2.18, allows context-dependent attackers to execute arbitrary code via vectors related to the (1) gethostbyname or (2) gethostbyname2 function, aka GHOST." [6]

Analysis

The following analysis is based on information in [6,7,8]. The number of bytes that can be overwritten is sizeof (char *), which is 4 bytes on a 32 bit machine, and 8 bytes on a 64 bit machine. In a calculation of the size needed to store certain data, the size of a char pointer is missing, resulting in array too small. Buffer over write is done by strcpy (continuous excursion). Qualys developed an attack on the Exim mail server, exploiting this vulnerability, as proof of concept. This attack uses an initial buffer over write to enlarge the number in the size field of a portion of memory that is available for the next allocation. This modification enables a subsequent over write that enables write-anything-anywhere, which in turn enables over writing Exim's Access Control Lists, which in turn enables arbitrary code execution.

Source Code

Code With Bug

1 2 3 4 5 6 7 /* calculate size incorrectly*/ size_needed = (sizeof (*host_addr)+ sizeof (*h_addr_ptrs)       + strlen (name) + 1); host_addr = (host_addr_t *) *buffer; h_addr_ptrs = (host_addr_list_t *)(( char *) host_addr       + sizeof (*host_addr)); hostname = (char *) h_addr_ptrs + sizeof (*h_addr_ptrs); resbuf->h_name = strcpy (hostname, name);

Code With Fix

1 2 3 4 5 6 7 /* calculate size incorrectly*/ size_needed = (sizeof (*host_addr) + sizeof (*h_addr_ptrs) + sizeof (*h_alias_ptr) + strlen (name) + 1); host_addr = (host_addr_t *) *buffer; h_addr_ptrs = (host_addr_list_t *)((char*) host_addr       + sizeof (*host_addr)); hostname = (char*) h_addr_ptrs + sizeof (*h_addr_ptrs); resbuf->h_name = strcpy (hostname, name);

CVE-2010-1773 (Chrome WebCore)

BF Taxonomy

Cause: Incorrect calculation, (off by one) leads to a wrong index

Attributes:
Access: Read
Boundary: Below
Location: Heap
Data Size: Little
Excursion: Discrete

Consequence: Information exposure, arbitrary code execution or program crash, leading to DoS

BF Description

"Incorrect calculation, (off by one) leads to a wrong index , where little data is read from the heap in a discrete excursion below the array boundary, which may be exploited for information exposure, arbitrary code execution or program crash, leading to DoS." [2]

CVE Description

"Off-by-one error in the toAlphabetic function in rendering/RenderListMarker.cpp in WebCore in WebKit before r59950, as used in Google Chrome before 5.0.375.70, allows remote attackers to obtain sensitive information, cause a denial of service (memory corruption and application crash), or possibly execute arbitrary code via vectors related to list markers for HTML lists, aka rdar problem 8009118." [9]

Analysis

The following analysis is based on information in [9,10,11,12,13,14,15,16].The software reads in a loop from an array, where the sequence of indices of array elements read is neither necessarily monotonic nor necessarily having a fixed distance between consecutive elements. That index should be the remainder obtained by dividing an integer by an integer. The software subtracts 1 from that remainder, which is wrong, and can result in the index being equal to -1, leading to reading from an address that is below the beginning of the array by 1. Consequences are mentioned in [10], and [16] includes "An off by one memory read out of bounds issue exists in WebKit's handling of HTML lists. Visiting a maliciously crafted website may lead to an unexpected application termination or the disclosure of the contents of memory."

Source Code

Code With Bug

1 2 3 4 5 6 7 if (type == AlphabeticSequence) { while ((numberShadow /= sequenceSize) > 0){ letters[lettersSize - ++length] = sequence[numberShadow % sequenceSize - 1]; } }

Code With Fix

1 2 3 4 5 6 7 if (type == AlphabeticSequence) { while ((numberShadow /= sequenceSize) > 0){ --numberShadow; letters[lettersSize - ++length] = sequence[numberShadow % sequenceSize]; } }

CVE-2015-2282

BF Taxonomy

Cause: Input not checked properly leads to wrong index

Attributes:
Access: Write
Boundary: Above
Location: Heap
Data Size: Some
Excursion: Continuous

Consequence: Arbitrary code execution or program crash, which can lead to DoS

BF Description

"Input not checked properly leads to wrong index, where some data is written to the heap in a continuous excursion, above the array boundary, which can cause arbitrary code execution or program crash, which can lead to DoS." [19]

CVE Description

"Stack-based buffer overflow in the LZC decompression implementation (CsObjectInt::CsDecomprLZC function in vpa106cslzc.cpp) in SAP MaxDB 7.5 and 7.6, Netweaver Application Server ABAP, Netweaver Application Server Java, Netweaver RFC SDK, GUI, RFC SDK, SAPCAR archive tool, and other products allows context-dependent attackers to cause a denial of service (crash) or possibly execute arbitrary code via unspecified vectors, aka SAP Security Note 2124806, 2121661, 2127995, and 2125316." [17]

Analysis

[17, 18, 19, 20] include sufficient information for describing this CVE using BF. Note that this buffer resides in the Heap even though it implements a stack data structure.

Source Code

Code With Bug	Code With Fix
Source Code Not Available	Source Code Not Available

CVE-2015-2278

BF Taxonomy

Cause: Input not checked properly leads to wrong index

Attributes:
Access: Read
Boundary: Above or Below
Data Size: Some
Excursion: Discrete

Consequence: DoS

BF Description

"Input not checked properly leads to wrong index, where some data is read in a discrete excursion, above or below the array boundary, which can lead to DoS ."

CVE Description

"The LZH decompression implementation (CsObjectInt::BuildHufTree function in vpa108csulzh.cpp) in SAP MaxDB 7.5 and 7.6, Netweaver Application Server ABAP, Netweaver Application Server Java, Netweaver RFC SDK, GUI, RFC SDK, SAPCAR archive tool, and other products allows context-dependent attackers to cause a denial of service (out-of-bounds read) via unspecified vectors, related to look-ups of non-simple codes, aka SAP Security Note 2124806, 2121661, 2127995, and 2125316." [21]

Analysis

[18, 21, 22] include sufficient information for describing this CVE using BF.

Source Code

Code With Bug	Code With Fix
Source Code Not Available	Source Code Not Available

CVE-2018-19842

BF Taxonomy

Cause: Boundary Not Checked Properly leads to Pointer Out of Range

Attributes:
Access: Read
Boundary: Above
Location: Stack
Data Size: Small (1 byte)
Excursion: Continuous

Consequence: Program Crash leading to DoS

BF Description

"Boundary Not Checked Properly leads to Pointer Out of Range, where small data is read in a continuous excursion, above the array boundary, which can lead to a Program Crash leading to DoS."

CVE Description

"getToken in libr/asm/p/asm_x86_nz.c in radare2 before 3.1.0 allows attackers to cause a denial of service (stack-based buffer over-read) via crafted x86 assembly data, as demonstrated by rasm2."

Analysis

In the assemble() function, char op[128] is defined, filled with the first 127 characters of the passed rasm2 command arguments, and terminated with 0 (for to make a valid C string). So, there is no problem there.

However, in the getToken() function, the part that skips spaces in str, does no check if the pointer bypasses the 0 terminator. The problem is only when the spaces go up to the end of str. If there were even only one character after the skipped spaces and before the end of str, the code works fine.

So, the 0 terminator is skipped only if while skipping spaces the last element in str is a space -- it could be only one space that is also the last one as in the Steps To Reproduce here.

The main fixes check if whether the pointer reaches or bypasses the end of the string (buffer) :

str[*begin]

str[*end]

if (*begin > strlen (str))

Note: There are still problems in the fixed code, as *begin and *end, are used before checking if they are null -- see lines 4298, 4307, 4311, 4318. If the assumption is they are never null, then why such a check is present in the while loops.

Source Code

Code With Bug

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 static x86newTokenType getToken(const char *str, size_t *begin, size_t *end) { while (begin && isspace ((ut8)str[*begin])) { ++(*begin); } if (!str[*begin]) { // null byte *end = *begin; return TT_EOF; } else if (isalpha ((ut8)str[*begin])) { // word token *end = *begin; while (end && isalnum ((ut8)str[*end])) { ++(*end); } return TT_WORD; } else if (isdigit ((ut8)str[*begin])) { // number token *end = *begin; while (end && isalnum ((ut8)str[*end])) { // accept alphanumeric characters, because hex. ++(*end);

Code With Fix

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 static x86newTokenType getToken(const char *str, size_t *begin, size_t *end) { if (*begin > strlen (str)) { return TT_EOF; } while (begin && str[*begin] && isspace ((ut8)str[*begin])) { ++(*begin); } if (!str[*begin]) { // null byte *end = *begin; return TT_EOF; } if (isalpha ((ut8)str[*begin])) { // word token *end = *begin; while (end && str[*end] && isalnum ((ut8)str[*end])) { ++(*end); } return TT_WORD; } if (isdigit ((ut8)str[*begin])) { // number token *end = *begin; while (end && isalnum ((ut8)str[*end])) { // accept alphanumeric characters, because hex. ++(*end);

References

[1] The MITRE Corporation, CVE Common Vulnerabilities and Exposures, CVE-2014-0160.

[2] Irena Bojanova, Paul E. Black, Yaacov Yesha, and Yan Wu, The Bugs Framework (BF): A Structured Approach to Express Bugs, QRS 2016, Vienna, Austria.

[3] S. Cassidy. Diagnosis of the OpenSSL Heartbleed, Mon 07 April 2014.

[4] WIKIPEDIA, The Free Encyclopedia, Heartbleed.

[5] The MITRE Corporation, CWE Common Weakness Enumeration, CWE-126 Buffer Over-read.

[6] The MITRE Corporation, CVE Common Vulnerabilities and Exposures, CVE-2015-0235.

[7] Openwall, bringing security into open environment, Qualys Security Advisory CVE-2015-0235, 27 January 2015.

[8] Qualys Security Advisory CVE-2015-0235.

[9] The MITRE Corporation, CVE Common Vulnerabilities and Exposures, CVE-2010-1773.

[10] Robin Gandhi, Buffer Overflow Semantic template CVE-2010-1773.

[11] Tracker, Issue 44955.

[12] chromium, Diff of /branches/WebKit/375/WebCore/rendering/RenderListMarker.cpp. Revision 48099.

[13] chromium, Contents of /branches/WebKit/375/WebCore/rendering/RenderListMarker.cpp. Revision 44321.

[14] chromium, Contents of /branches/WebKit/375/WebCore/rendering/RenderListMarker.cpp. Revision 48100.

[15] webkit, Fix for Crash in WebCore::toAlphabetic() while running MangleMe -and corresponding- https://bugs.webkit.org/show_bug.cgi?id=39508. Reviewed by Darin Adler.

[16] Hat Bugzilla - Bug 596500- (CVE-2010-1773) CVE-2010-1773 WebKit: off-by-one memory read out of bounds vulnerability in handling of HTML lists.

[17] The MITRE Corporation, CVE Common Vulnerabilities and Exposures, CVE-2015-2282.

[18] Core Security, SAP LZC LZH Compression Multiple Vulnerabilities. (referenced in [17] and [21]).

[19] SourceArchive, SAP AG, vpa106cslzc.cpp (reference 12 in [18]).

[20] SourceArchive, SAP AG, hpa106cslzc.h.

[21] The MITRE Corporation, CVE Common Vulnerabilities and Exposures, CVE-2015-2278.

[22] SourceArchive, SAP AG, vpa108csulzh.cpp. (reference 13 in [18]).

[23] sourceware, blobdiff.