1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
# Tag Image File Format, from Daniel Quinlan (quinlan@yggdrasil.com)
# The second word of TIFF files is the TIFF version number, 42, which has
# never changed. The TIFF specification recommends testing for it.
0 string MM\x00\x2a TIFF image data, big-endian
0 string II\x2a\x00 TIFF image data, little-endian
# PNG [Portable Network Graphics, or "PNG's Not GIF"] images
# (Greg Roelofs, newt@uchicago.edu)
# (Albert Cahalan, acahalan@cs.uml.edu)
#
# 137 P N G \r \n ^Z \n [4-byte length] H E A D [HEAD data] [HEAD crc] ...
#
0 string \x89PNG\x0d\x0a\x1a\x0a PNG image
>16 belong <1 invalid
>16 belong >10000 invalid
>20 belong <1 invalid
>20 belong >10000 invalid
>16 belong x \b, %ld x
>20 belong x %ld,
>24 byte x %d-bit
>25 byte 0 grayscale,
>25 byte 2 \b/color RGB,
>25 byte 3 colormap,
>25 byte 4 gray+alpha,
>25 byte 6 \b/color RGBA,
#>26 byte 0 deflate/32K,
>28 byte 0 non-interlaced
>28 byte 1 interlaced
# GIF
0 string GIF8 GIF image data
>4 string 7a \b, version "8%s",
>4 string 9a \b, version "8%s",
>6 leshort >0 %hd x
>8 leshort >0 %hd
#>10 byte &0x80 color mapped,
#>10 byte&0x07 =0x00 2 colors
#>10 byte&0x07 =0x01 4 colors
#>10 byte&0x07 =0x02 8 colors
#>10 byte&0x07 =0x03 16 colors
#>10 byte&0x07 =0x04 32 colors
#>10 byte&0x07 =0x05 64 colors
#>10 byte&0x07 =0x06 128 colors
#>10 byte&0x07 =0x07 256 colors
# PC bitmaps (OS/2, Windows BMP files) (Greg Roelofs, newt@uchicago.edu)
0 string BM
>14 leshort 12 PC bitmap, OS/2 1.x format
>>18 lelong <1 invalid
>>18 lelong >1000000 invalid
>>18 leshort x \b, %d x
>>20 lelong <1 invalid
>>20 lelong >1000000 invalid
>>20 leshort x %d
>14 leshort 64 PC bitmap, OS/2 2.x format
>>18 lelong <1 invalid
>>18 lelong >1000000 invalid
>>18 leshort x \b, %d x
>>20 lelong <1 invalid
>>20 lelong >1000000 invalid
>>20 leshort x %d
>14 leshort 40 PC bitmap, Windows 3.x format
>>18 lelong <1 invalid
>>18 lelong >1000000 invalid
>>18 lelong x \b, %d x
>>22 lelong <1 invalid
>>22 lelong >1000000 invalid
>>22 lelong x %d x
>>28 lelong <1 invalid
>>28 lelong >1000000 invalid
>>28 leshort x %d
>14 leshort 128 PC bitmap, Windows NT/2000 format
>>18 lelong >1000000 invalid
>>18 lelong <1 invalid
>>18 lelong x \b, %d x
>>22 lelong <1 invalid
>>22 lelong >1000000 invalid
>>22 lelong x %d x
>>28 lelong <1 invalid
>>28 lelong >1000000 invalid
>>28 leshort x %d
#------------------------------------------------------------------------------
# JPEG images
# SunOS 5.5.1 had
#
# 0 string \377\330\377\340 JPEG file
# 0 string \377\330\377\356 JPG file
#
# both of which turn into "JPEG image data" here.
#
0 belong 0xffd8ffe0 JPEG image data, JFIF standard
>6 string !JFIF invalid
# The following added by Erik Rossen <rossen@freesurf.ch> 1999-09-06
# in a vain attempt to add image size reporting for JFIF. Note that these
# tests are not fool-proof since some perfectly valid JPEGs are currently
# impossible to specify in magic(4) format.
# First, a little JFIF version info:
>11 byte x \b %d.
>12 byte x \b%02d
# Next, the resolution or aspect ratio of the image:
#>>13 byte 0 \b, aspect ratio
#>>13 byte 1 \b, resolution (DPI)
#>>13 byte 2 \b, resolution (DPCM)
#>>4 beshort x \b, segment length %d
# Next, show thumbnail info, if it exists:
>18 byte !0 \b, thumbnail %dx
>>19 byte x \b%d
0 belong 0xffd8ffe1 JPEG image data, EXIF standard
# EXIF moved down here to avoid reporting a bogus version number,
# and EXIF version number printing added.
# - Patrik R=E5dman <patrik+file-magic@iki.fi>
>6 string !Exif invalid
# Look for EXIF IFD offset in IFD 0, and then look for EXIF version tag in EXIF IFD.
# All possible combinations of entries have to be enumerated, since no looping
# is possible. And both endians are possible...
# The combinations included below are from real-world JPEGs.
# Little-endian
>12 string II
# IFD 0 Entry #5:
>>70 leshort 0x8769
# EXIF IFD Entry #1:
>>>(78.l+14) leshort 0x9000
>>>>(78.l+23) byte x %c
>>>>(78.l+24) byte x \b.%c
>>>>(78.l+25) byte !0x30 \b%c
# IFD 0 Entry #9:
>>118 leshort 0x8769
# EXIF IFD Entry #3:
>>>(126.l+38) leshort 0x9000
>>>>(126.l+47) byte x %c
>>>>(126.l+48) byte x \b.%c
>>>>(126.l+49) byte !0x30 \b%c
# IFD 0 Entry #10
>>130 leshort 0x8769
# EXIF IFD Entry #3:
>>>(138.l+38) leshort 0x9000
>>>>(138.l+47) byte x %c
>>>>(138.l+48) byte x \b.%c
>>>>(138.l+49) byte !0x30 \b%c
# EXIF IFD Entry #4:
>>>(138.l+50) leshort 0x9000
>>>>(138.l+59) byte x %c
>>>>(138.l+60) byte x \b.%c
>>>>(138.l+61) byte !0x30 \b%c
# EXIF IFD Entry #5:
>>>(138.l+62) leshort 0x9000
>>>>(138.l+71) byte x %c
>>>>(138.l+72) byte x \b.%c
>>>>(138.l+73) byte !0x30 \b%c
# IFD 0 Entry #11
>>142 leshort 0x8769
# EXIF IFD Entry #3:
>>>(150.l+38) leshort 0x9000
>>>>(150.l+47) byte x %c
>>>>(150.l+48) byte x \b.%c
>>>>(150.l+49) byte !0x30 \b%c
# EXIF IFD Entry #4:
>>>(150.l+50) leshort 0x9000
>>>>(150.l+59) byte x %c
>>>>(150.l+60) byte x \b.%c
>>>>(150.l+61) byte !0x30 \b%c
# EXIF IFD Entry #5:
>>>(150.l+62) leshort 0x9000
>>>>(150.l+71) byte x %c
>>>>(150.l+72) byte x \b.%c
>>>>(150.l+73) byte !0x30 \b%c
# Big-endian
>12 string MM
# IFD 0 Entry #9:
>>118 beshort 0x8769
# EXIF IFD Entry #1:
>>>(126.L+14) beshort 0x9000
>>>>(126.L+23) byte x %c
>>>>(126.L+24) byte x \b.%c
>>>>(126.L+25) byte !0x30 \b%c
# EXIF IFD Entry #3:
>>>(126.L+38) beshort 0x9000
>>>>(126.L+47) byte x %c
>>>>(126.L+48) byte x \b.%c
>>>>(126.L+49) byte !0x30 \b%c
# IFD 0 Entry #10
>>130 beshort 0x8769
# EXIF IFD Entry #3:
>>>(138.L+38) beshort 0x9000
>>>>(138.L+47) byte x %c
>>>>(138.L+48) byte x \b.%c
>>>>(138.L+49) byte !0x30 \b%c
# EXIF IFD Entry #5:
>>>(138.L+62) beshort 0x9000
>>>>(138.L+71) byte x %c
>>>>(138.L+72) byte x \b.%c
>>>>(138.L+73) byte !0x30 \b%c
# IFD 0 Entry #11
>>142 beshort 0x8769
# EXIF IFD Entry #4:
>>>(150.L+50) beshort 0x9000
>>>>(150.L+59) byte x %c
>>>>(150.L+60) byte x \b.%c
>>>>(150.L+61) byte !0x30 \b%c
# Here things get sticky. We can do ONE MORE marker segment with
# indirect addressing, and that's all. It would be great if we could
# do pointer arithemetic like in an assembler language. Christos?
# And if there was some sort of looping construct to do searches, plus a few
# named accumulators, it would be even more effective...
# At least we can show a comment if no other segments got inserted before:
>(4.S+5) byte 0xFE
>>(4.S+8) string >\0 \b, comment: "%s"
# FIXME: When we can do non-byte counted strings, we can use that to get
# the string's count, and fix Debian bug #283760
#>(4.S+5) byte 0xFE \b, comment
#>>(4.S+6) beshort x \b length=%d
#>>(4.S+8) string >\0 \b, "%s"
# Or, we can show the encoding type (I've included only the three most common)
# and image dimensions if we are lucky and the SOFn (image segment) is here:
>(4.S+5) byte 0xC0 \b, baseline
>>(4.S+6) byte x \b, precision %d
>>(4.S+7) beshort x \b, %dx
>>(4.S+9) beshort x \b%d
>(4.S+5) byte 0xC1 \b, extended sequential
>>(4.S+6) byte x \b, precision %d
>>(4.S+7) beshort x \b, %dx
>>(4.S+9) beshort x \b%d
>(4.S+5) byte 0xC2 \b, progressive
>>(4.S+6) byte x \b, precision %d
>>(4.S+7) beshort x \b, %dx
>>(4.S+9) beshort x \b%d
# I've commented-out quantisation table reporting. I doubt anyone cares yet.
#>(4.S+5) byte 0xDB \b, quantisation table
#>>(4.S+6) beshort x \b length=%d
#>14 beshort x \b, %d x
#>16 beshort x \b %d
0 string M88888888888888888888888888 Binwalk logo, ASCII art (Toph){offset-adjust:-50}
>27 string !8888888888\n invalid