Difference between pages "Prefetch" and "Windows Prefetch File Format"

From ForensicsWiki
(Difference between pages)
Jump to: navigation, search
 
(Volume information - version 17)
 
Line 1: Line 1:
{{Expand}}
+
{{expand}}
Windows Prefetch files, introduced in [[Windows|Windows XP]], are designed to speed up the application startup process. Prefetch files contain the name of the executable, a Unicode list of DLLs used by that executable, a count of how many times the executable has been run, and a timestamp indicating the last time the program was run. Although Prefetch is present in Windows 2003, by default it is only enabled for boot prefetching. The feature is also found in [[Windows|Windows Vista]], where it has been augmented with [[SuperFetch]], [[ReadyBoot]], and [[ReadyBoost]]. For SSD drives Prefetch is disabled by default [http://blogs.msdn.com/b/e7/archive/2009/05/05/support-and-q-a-for-solid-state-drives-and.aspx].
+
  
Up to 128 Prefetch files are stored in the <tt>%SystemRoot%\Prefetch</tt> directory [http://blogs.msdn.com/ryanmy/archive/2005/05/25/421882.aspx]. Each file in that directory should contain the name of the application, a dash, and then an eight character hash of the location from which that application was run, and a <tt>.pf</tt> extension. The filenames should be all uppercase except for the extension. The format of hashes is not known. A sample filename for [[md5deep]] would look like: <tt>MD5DEEP.EXE-4F89AB0C.pf</tt>. If an application is run from two different locations on the drive (i.e. the user runs <tt>C:\md5deep.exe</tt> and then <tt>C:\Apps\Hashing\md5deep.exe</tt>), there will be two different prefetch files in the Prefetch folder.
+
A Windows Prefetch file consists of one file header and multiple file sections with different content. Not all content has an obvious forensic value.
  
== File format ==
+
As far as have been possible to ascertain, there is no public description of the format. The description below has been synthesised from examination
Each Prefetch file has a 4-byte signature (at offset 4) "SCCA" (or in hexadecimal notation 0x53 0x43 0x43 0x41). The signature is assumed to be preceded by a 4-byte format version indicator:
+
of multiple prefetch files.
* 17 (0x00000011) for [[Windows XP]] and [[Windows 2003]]
+
* 23 (0x00000017) for [[Windows Vista]], [[Windows 2008]], [[Windows 7]] and [[Windows 2012]] (note Windows 2012 has not been confirmed)
+
* 26 (0x0000001a) for [[Windows 8|Windows 8.1]] (note this could be Windows 8 as well but has not been confirmed)
+
  
For more information about the file format see: [[Windows Prefetch File Format]]
+
== Characteristics ==
 +
Integer values are stored in little-endian.
  
== Metadata ==
+
Strings are stored as [http://en.wikipedia.org/wiki/UTF-16/UCS-2 UTF-16 little-endian] without a byte-order-mark (BOM).
The Prefetch file contains various metadata.
+
* The executable's name, up to 29 characters.
+
* The run count, or number of times the application has been run.
+
* Volume related information, like volume path and volume serial number.
+
* The size of the Prefetch file (sometimes referred to as end of file (EOF)).
+
* The files and directories that were used doing the application's start-up.
+
  
=== Timestamps ===
+
Timestamps are stored as [http://msdn2.microsoft.com/en-us/library/ms724284.aspx Windows FILETIME] in UTC.
The Prefetch file contains 2 types of timestamps
+
* The time when the application was last ran (executed). Version 26 of the Prefetch format maintains 7 previous last run times.
+
* The volume creation time (part of the volume information).
+
  
The creation date of the Prefetch file indicates the first time the application was executed. Both the modification date of the Prefetch file and the embedded Last Run timestamp indicate the last time the application was executed.
+
== File header ==
  
== Prefetch hash ==
+
{| class="wikitable"
There are multiple known hashing functions to be used for prefetch file filename hashing, namely:
+
|-
* SCCA XP hash function; used on Windows XP and Windows 2003
+
! Field
* SCCA Vista hash function; used on Windows Vista
+
! Offset
* SCCA 2008 hash function; used on Windows 2008, Windows 7, (possibly: Windows 2012) and Windows 8 (including 8.1)
+
! Length
 +
! Type
 +
! Notes
 +
|-
 +
| H1
 +
| 0x0000
 +
| 4
 +
| DWORD
 +
| Format version (see format version section below)
 +
|-
 +
| H2
 +
| 0x0004
 +
| 4
 +
| DWORD
 +
| Signature 'SCCA' (or in hexadecimal representation 0x53 0x43 0x43 0x4)
 +
|-
 +
| H3
 +
| 0x0008
 +
| 4
 +
| DWORD?
 +
| Unknown - Values observed: 0x0F - Windows XP, 0x11 - Windows 7, Windows 8.1
 +
|-
 +
| H4
 +
| 0x000C
 +
| 4
 +
| DWORD
 +
| Prefetch file size (or length) (sometimes referred to as End of File (EOF)).
 +
|-
 +
| H5
 +
|0x0010
 +
| 60
 +
| USTR
 +
| The name of the (original) executable as a Unicode (UTF-16 litte-endian string), up to 29 characters and terminated by an end-of-string character (U+0000). This name should correspond with the one in the prefetch file filename.
 +
|-
 +
| H6
 +
|0x004C
 +
|4
 +
|DWORD
 +
|The prefetch hash. This hash value should correspond with the one in the prefetch file filename.
 +
|-
 +
| H7
 +
|0x0050
 +
|4
 +
|?
 +
| Unknown (flags)? Values observed: 0 for almost all prefetch files (XP); 1 for NTOSBOOT-B00DFAAD.pf (XP)
 +
|-
 +
|}
  
=== SCCA XP hash function ===
+
=== Format version ===
A Python implementation of the SCCA XP hash function:
+
  
<pre>
+
{| class="wikitable"
def ssca_xp_hash_function(filename):
+
|-
    hash_value = 0
+
! Value
    for character in filename:
+
! Windows version
        hash_value = ((hash_value * 37) + ord(character)) % 0x100000000
+
|-
        hash_value = (hash_value * 314159269) % 0x100000000
+
| 17 (0x11)
        if hash_value > 0x80000000:
+
| Windows XP, Windows 2003
            hash_value = 0x100000000 - hash_value
+
|-
 +
| 23 (0x17)
 +
| Windows Vista, Windows 7
 +
|-
 +
| 26 (0x1a)
 +
| Windows 8.1 (note this could be Windows 8 as well but has not been confirmed)
 +
|-
 +
|}
  
    return (abs(hash_value) % 1000000007) % 0x100000000
+
=== File information - version 17 ===
</pre>
+
  
=== SCCA Vista hash function ===
+
The following part of the file header is version dependent. It is sometime considered part of the file header. Below the structure for format version 17.
A Python implementation of the SCCA Vista hash function:
+
  
<pre>
+
{| class="wikitable"
def ssca_vista_hash_function(filename):
+
|-
    hash_value = 314159
+
! Field
    for character in filename:
+
! Offset
        hash_value = ((hash_value * 37) + ord(character)) % 0x100000000
+
! Length
    return hash_value
+
! Type
</pre>
+
! Notes
 +
|-
 +
| H8
 +
| 0x0054
 +
| 4
 +
| DWORD
 +
| The offset to section A. The offset is relative from the start of the file.
 +
|-
 +
| H9
 +
| 0x0058
 +
| 4
 +
| DWORD
 +
| The number of entries in section A.
 +
|-
 +
| H10
 +
| 0x005C
 +
| 4
 +
| DWORD
 +
| The offset to section B. The offset is relative from the start of the file.
 +
|-
 +
| H11
 +
| 0x0060
 +
| 4
 +
| DWORD
 +
| The number of entries in section B.
 +
|-
 +
| H12
 +
| 0x0064
 +
| 4
 +
| DWORD
 +
| The offset to section C. The offset is relative from the start of the file.
 +
|-
 +
| H13
 +
| 0x0068
 +
| 4
 +
| DWORD
 +
| Length of section C.
 +
|-
 +
| H14
 +
| 0x006C
 +
| 4
 +
| DWORD
 +
| Offset to section D. The offset is relative from the start of the file.
 +
|-
 +
| H15
 +
| 0x0070
 +
| 4
 +
| DWORD
 +
| Unknown ? (Previously opted: Probably the number of entries in the D section header.)
 +
|-
 +
| H16
 +
| 0x0074
 +
| 4
 +
| DWORD
 +
| Unknown ? (Previously opted: Length of section D)
 +
|-
 +
| H17
 +
| 0x0078
 +
| 8
 +
| FTIME
 +
| Latest execution time (or run time) of executable (FILETIME)
 +
|-
 +
| H18
 +
| 0x0080
 +
| 16
 +
| ?
 +
| Unknown ? Possibly structured as 4 DWORD. Observed values: /0x00000000 0x00000000 0x00000000 0x00000000/, /0x47868c00 0x00000000 0x47860c00 0x00000000/ (don't exclude the possibility here that this is remnant data)
 +
|-
 +
| H19
 +
| 0x0090
 +
| 4
 +
| DWORD
 +
| Execution counter (or run count)
 +
|-
 +
| H20
 +
| 0x0094
 +
| 4
 +
| DWORD?
 +
| Unknown ? Observed values: 1, 2, 3, 4, 5, 6 (XP)
 +
|-
 +
|}
  
=== SCCA 2008 hash function ===
+
It's worth noting that the name of a carved prefetch file can be restored using the information in field H5 and H6, and its size can be determined by field H4.
A Python implementation of the SCCA 2008 hash function:
+
  
<pre>
+
== Section A ==
def ssca_2008_hash_function(filename):
+
This section contains an array with 20 byte (version 17) or 32 byte (version 23 and 26) entry records.
    hash_value = 314159
+
    filename_index = 0
+
    filename_length = len(filename)
+
    while filename_index + 8 < filename_length:
+
        character_value = ord(filename[filename_index + 1]) * 37
+
        character_value += ord(filename[filename_index + 2])
+
        character_value *= 37
+
        character_value += ord(filename[filename_index + 3])
+
        character_value *= 37
+
        character_value += ord(filename[filename_index + 4])
+
        character_value *= 37
+
        character_value += ord(filename[filename_index + 5])
+
        character_value *= 37
+
        character_value += ord(filename[filename_index + 6])
+
        character_value *= 37
+
        character_value += ord(filename[filename_index]) * 442596621
+
        character_value += ord(filename[filename_index + 7])
+
        hash_value = ((character_value - (hash_value * 803794207)) % 0x100000000)
+
        filename_index += 8
+
  
    while filename_index < filename_length:
+
The actual format and usage of these entry records is currently not known.
      hash_value = (((37 * hash_value) + ord(filename[filename_index])) % 0x100000000)
+
      filename_index += 1
+
  
    return hash_value
+
== Section B ==
</pre>
+
This section contains an array with 12 byte (version 17, 23 and 26) entry records.
  
== Registry Keys ==
+
The actual format and usage of these entry records is currently not known.
<pre>
+
Key: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management\PrefetchParameters
+
</pre>
+
  
The EnablePrefetcher Registry value can be used to disable prefetch.
+
== Section C ==
 +
This section contains an array of UTF-16 little-endian formatted strings with end-of-string characters (U+0000).
  
== See Also ==
+
At the end of the section there seems to be alignment padding that can contain remnant values.
* [[Windows Prefetch File Format]]
+
* [[SuperFetch]]
+
* [[Prefetch XML]]
+
* [[Windows]]
+
  
== External Links ==
+
== Section D - Volume information (block) ==
* [http://msdn.microsoft.com/msdnmag/issues/01/12/XPKernel/default.aspx More detail from Microsoft]
+
* [http://en.wikipedia.org/wiki/Prefetcher Wikipedia Prefetcher]
+
* [http://msdn.microsoft.com/en-us/library/ms940847(v=winembedded.5).aspx MSDN: Disabling Prefetch]
+
* [http://www.microsoft.com/whdc/driver/kernel/XP_kernel.mspx Kernel Enhancements for Windows XP], by [[Microsoft]], January 13, 2003 (Microsoft's description of Prefetch when Windows XP was introduced)
+
* [http://blogs.msdn.com/b/ryanmy/archive/2005/05/25/421882.aspx Misinformation and the The Prefetch Flag], MSDN Blogs, May 25, 2005
+
* [http://windowsir.blogspot.ch/2005/07/prefetch-file-metadata.html Prefetch file metadata], by [[Harlan Carvey]], July 13, 2005
+
* [http://windowsir.blogspot.ch/2006/04/prefetch-files-revisited.html Prefetch files, revisited], by [[Harlan Carvey]], April 13, 2006
+
* [http://blogs.msdn.com/b/e7/archive/2009/05/05/support-and-q-a-for-solid-state-drives-and.aspx Support and Q&A for Solid-State Drives], by Steven Sinofsky, May 5, 2009
+
* [http://computer-forensics.sans.org/blog/2009/08/05/de-mystifying-defrag-identifying-when-defrag-has-been-used-for-anti-forensics-part-1-windows-xp/ De-mystifying Defrag: Identifying When Defrag Has Been Used for Anti-Forensics (Part 1 - Windows XP)], by [[Chad Tilbury]], August 5, 2009
+
* [http://www.swiftforensics.com/2010/04/the-windows-prefetchfile.html Windows Prefetch File (old blog entry from 42 LLC)], by [[Yogesh Khatri]], April 14, 2010
+
* [http://www.dfinews.com/articles/2010/12/decoding-prefetch-files-forensic-purposes-part-1 Decoding Prefetch Files for Forensic Purposes: Part 1], by [[Mark Wade]], December 8, 2010
+
* [http://crucialsecurityblog.harris.com/2011/04/11/prefetch-files-at-face-value/ Prefetch Files at Face Value], by [[Mark Wade]], April 11, 2011
+
* [http://kitrap08.blogspot.hk/2011/07/windows-logical-prefetcher.html Windows Logical Prefetcher], TTS blog, July 30, 2011 (article is in Russian)
+
* [http://labit.in/pliki-prefetch-w-windows/ Prefetch i niedokładny licznik] by Paweł Hałdrzyński, August 20, 2011 (article in Polish; uncertain about the year of publication)
+
* [http://windowsir.blogspot.ch/2012/03/prefetch-analysis-revisited.html Prefetch Analysis, Revisited], by [[Harlan Carvey]], March 8, 2012
+
* [http://windowsir.blogspot.ch/2012/03/prefetch-analysis-revisitedagain.html Prefetch Analysis, Revisited...Again...], by [[Harlan Carvey]], March 15, 2012
+
* [http://www.hexacorn.com/blog/2012/06/13/prefetch-hash-calculator-a-hash-lookup-table-xpvistaw7w2k3w2k8/ Prefetch Hash Calculator + a hash lookup table xp/vista/w7/w2k3/w2k8], Hexacorn blog, June 13, 2012
+
* [http://www.hexacorn.com/blog/2012/10/29/prefetch-file-names-and-unc-paths/ Prefetch file names and UNC paths], Hexacorn blog, October 29, 2012
+
* [http://journeyintoir.blogspot.ch/2012/12/ntosboot-prefetch-file.html NTOSBOOT Prefetch File], by [[Corey Harrell]], December 5, 2012
+
* [http://www.invoke-ir.com/2013/09/whats-new-in-prefetch-for-windows-8.html What's New in the Prefetch for Windows 8??], by Jared Atkinson, September 21, 2013
+
* [http://www.swiftforensics.com/2013/10/windows-prefetch-pf-files.html?m=1 Windows Prefetch (.PF) files], by [[Yogesh Khatri]], October 21, 2013
+
  
== Tools ==
+
Section D contains one or more subsections. The number is (most likely) determined by the DWORD at file offset 0x0070. Each subsection refers to directories on an identified volume.
  
=== Commercial ===
+
In this section, all offsets are assumed to be counted from the start of the D section.
  
=== Free - Non Open Source ===
+
=== Volume information - version 17 ===
* [http://www.woanware.co.uk/forensics/prefetchforensics.html PrefetchForensics], PrefetchForensics is an application to extract information from Windows Prefetch files
+
The following values are version dependent. Below the structure for format version 17.
* [http://redwolfcomputerforensics.com/index.php?option=com_content&task=view&id=42&Itemid=55 Prefetch-Parser], Parse the prefetch files and display information
+
* [http://www.mitec.cz/wfa.html Windows File Analyzer] - Parses Prefetch files, thumbnail databases, shortcuts, index.dat files, and the recycle bin
+
* [http://www.tzworks.net/prototype_page.php?proto_id=1 Windows Prefetch Parser (pf)], Free tool that can be run on Windows, Linux or Mac OS-X
+
  
=== Open Source ===
+
{| class="wikitable"
* [https://code.google.com/p/prefetch-tool/ prefetch-tool], Script to extract information from windows prefetch folder
+
|-
 +
! Field
 +
! Offset
 +
! Length
 +
! Type
 +
! Notes
 +
|-
 +
| DH1
 +
| +0x0000
 +
| 4
 +
| DWORD
 +
| Offset to volume string (Unicode, terminated by U+0000)
 +
|-
 +
| DH2
 +
| +0x0004
 +
| 4
 +
| DWORD
 +
| Length of volume string (nr of characters, including terminating U+0000)
 +
|-
 +
| DH3
 +
| +0x0008
 +
| 8
 +
| FILETIME
 +
| Volume creation time.
 +
|-
 +
| DH4
 +
| +0x0010
 +
| 4
 +
| DWORD
 +
| Volume serial number of volume indicated by volume string
 +
|-
 +
| DH5
 +
| +0x0014
 +
| 4
 +
| DWORD
 +
| ? Offset to section DHS1
 +
|-
 +
| DH6
 +
| +0x0018
 +
| 4
 +
| DWORD
 +
| ? Length of section DHS1 (in bytes)
 +
|-
 +
| DH7
 +
| +0x001C
 +
| 4
 +
| DWORD
 +
| ? Offset to section DHS2
 +
|-
 +
| DH8
 +
| +0x0020
 +
| 4
 +
| DWORD
 +
| ? Nr of strings in section DHS2
 +
|-
 +
| ?
 +
| +0x0024
 +
| ?
 +
| ?
 +
| ? additional 28 bytes (includes one timestamp?)
 +
|}
 +
 
 +
If all the executables and libraries referenced in the C section are from one single disk volume, there will be only one section in the D section. If multiple volumes are referenced by section C, section D will contain multiple sections.  (A simple way to force this situation is to copy, say, NOTEPAD.EXE to a USB drive, and start it from that volume. The corresponding prefetch file will have one D header referring to, e.g. \DEVICE\HARDDISK1\DP(1)0-0+4 (the USB drive), and one to, e.g. \DEVICE\HARDDISKVOLUME1\ (where the .DLLs and other support files were found).
 +
 
 +
== See Also ==
 +
* [[Prefetch]]
 +
 
 +
== External Links ==
 +
* [https://googledrive.com/host/0B3fBvzttpiiSbl9XZGZzQ05hZkU/Windows%20Prefetch%20File%20(PF)%20format.pdf Windows Prefetch File (PF) format], by the [[libssca|libssca project]]

Revision as of 04:31, 2 November 2013

Information icon.png

Please help to improve this article by expanding it.
Further information might be found on the discussion page.

A Windows Prefetch file consists of one file header and multiple file sections with different content. Not all content has an obvious forensic value.

As far as have been possible to ascertain, there is no public description of the format. The description below has been synthesised from examination of multiple prefetch files.

Characteristics

Integer values are stored in little-endian.

Strings are stored as UTF-16 little-endian without a byte-order-mark (BOM).

Timestamps are stored as Windows FILETIME in UTC.

File header

Field Offset Length Type Notes
H1 0x0000 4 DWORD Format version (see format version section below)
H2 0x0004 4 DWORD Signature 'SCCA' (or in hexadecimal representation 0x53 0x43 0x43 0x4)
H3 0x0008 4 DWORD? Unknown - Values observed: 0x0F - Windows XP, 0x11 - Windows 7, Windows 8.1
H4 0x000C 4 DWORD Prefetch file size (or length) (sometimes referred to as End of File (EOF)).
H5 0x0010 60 USTR The name of the (original) executable as a Unicode (UTF-16 litte-endian string), up to 29 characters and terminated by an end-of-string character (U+0000). This name should correspond with the one in the prefetch file filename.
H6 0x004C 4 DWORD The prefetch hash. This hash value should correspond with the one in the prefetch file filename.
H7 0x0050 4 ? Unknown (flags)? Values observed: 0 for almost all prefetch files (XP); 1 for NTOSBOOT-B00DFAAD.pf (XP)

Format version

Value Windows version
17 (0x11) Windows XP, Windows 2003
23 (0x17) Windows Vista, Windows 7
26 (0x1a) Windows 8.1 (note this could be Windows 8 as well but has not been confirmed)

File information - version 17

The following part of the file header is version dependent. It is sometime considered part of the file header. Below the structure for format version 17.

Field Offset Length Type Notes
H8 0x0054 4 DWORD The offset to section A. The offset is relative from the start of the file.
H9 0x0058 4 DWORD The number of entries in section A.
H10 0x005C 4 DWORD The offset to section B. The offset is relative from the start of the file.
H11 0x0060 4 DWORD The number of entries in section B.
H12 0x0064 4 DWORD The offset to section C. The offset is relative from the start of the file.
H13 0x0068 4 DWORD Length of section C.
H14 0x006C 4 DWORD Offset to section D. The offset is relative from the start of the file.
H15 0x0070 4 DWORD Unknown ? (Previously opted: Probably the number of entries in the D section header.)
H16 0x0074 4 DWORD Unknown ? (Previously opted: Length of section D)
H17 0x0078 8 FTIME Latest execution time (or run time) of executable (FILETIME)
H18 0x0080 16  ? Unknown ? Possibly structured as 4 DWORD. Observed values: /0x00000000 0x00000000 0x00000000 0x00000000/, /0x47868c00 0x00000000 0x47860c00 0x00000000/ (don't exclude the possibility here that this is remnant data)
H19 0x0090 4 DWORD Execution counter (or run count)
H20 0x0094 4 DWORD? Unknown ? Observed values: 1, 2, 3, 4, 5, 6 (XP)

It's worth noting that the name of a carved prefetch file can be restored using the information in field H5 and H6, and its size can be determined by field H4.

Section A

This section contains an array with 20 byte (version 17) or 32 byte (version 23 and 26) entry records.

The actual format and usage of these entry records is currently not known.

Section B

This section contains an array with 12 byte (version 17, 23 and 26) entry records.

The actual format and usage of these entry records is currently not known.

Section C

This section contains an array of UTF-16 little-endian formatted strings with end-of-string characters (U+0000).

At the end of the section there seems to be alignment padding that can contain remnant values.

Section D - Volume information (block)

Section D contains one or more subsections. The number is (most likely) determined by the DWORD at file offset 0x0070. Each subsection refers to directories on an identified volume.

In this section, all offsets are assumed to be counted from the start of the D section.

Volume information - version 17

The following values are version dependent. Below the structure for format version 17.

Field Offset Length Type Notes
DH1 +0x0000 4 DWORD Offset to volume string (Unicode, terminated by U+0000)
DH2 +0x0004 4 DWORD Length of volume string (nr of characters, including terminating U+0000)
DH3 +0x0008 8 FILETIME Volume creation time.
DH4 +0x0010 4 DWORD Volume serial number of volume indicated by volume string
DH5 +0x0014 4 DWORD  ? Offset to section DHS1
DH6 +0x0018 4 DWORD  ? Length of section DHS1 (in bytes)
DH7 +0x001C 4 DWORD  ? Offset to section DHS2
DH8 +0x0020 4 DWORD  ? Nr of strings in section DHS2
 ? +0x0024  ?  ?  ? additional 28 bytes (includes one timestamp?)

If all the executables and libraries referenced in the C section are from one single disk volume, there will be only one section in the D section. If multiple volumes are referenced by section C, section D will contain multiple sections. (A simple way to force this situation is to copy, say, NOTEPAD.EXE to a USB drive, and start it from that volume. The corresponding prefetch file will have one D header referring to, e.g. \DEVICE\HARDDISK1\DP(1)0-0+4 (the USB drive), and one to, e.g. \DEVICE\HARDDISKVOLUME1\ (where the .DLLs and other support files were found).

See Also

External Links