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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
|
.\"
.\" Copyright (c) 2002 Poul-Henning Kamp
.\" Copyright (c) 2002 Networks Associates Technology, Inc.
.\" All rights reserved.
.\"
.\" This software was developed for the FreeBSD Project by Poul-Henning Kamp
.\" and NAI Labs, the Security Research Division of Network Associates, Inc.
.\" under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
.\" DARPA CHATS research program.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd October 19, 2002
.Os
.Dt GBDE 4
.Sh NAME
.Nm gbde
.Nd Geom Based Disk Encryption
.Sh SYNOPSIS
.Cd "options GEOM_BDE"
.Sh DESCRIPTION
.Bf -symbolic
NOTICE:
Please be aware that this code has not yet received much review
and analysis by qualified cryptographers and therefore should be considered
a slightly suspect experimental facility.
.Pp
We cannot at this point guarantee that the on-disk format will not change
in response to reviews or bug-fixes, so potential users are advised to
be prepared that
.Xr dump 8 Ns / Ns Xr restore 8
based migrations may be called for in the future.
.Ef
.Pp
The objective of this facility is to provide a high degree of
denial of access to the contents of a
.Dq cold
storage device.
.Pp
Be aware that if the computer is compromised while up and running
.Em and
the storage device is actively attached and opened with a valid
pass-phrase, this facility offers no protection or denial of access
to the contents of the storage device.
.Pp
If, on the other hand, the device is
.Dq cold ,
it should present an formidable
challenge for an attacker to gain access to the contents in the absence of
a valid pass-phrase.
.Pp
Four cryptographic barriers must be passed to gain access to the data,
and only a valid pass-phrase will yield this access.
.Pp
When the pass-phrase is entered, it is hashed with SHA2 into a 512 bit
.Dq key-material .
This is a way of producing cryptographic usable keys from a typically
.No all- Ns Tn ASCII
pass-phrase of an unpredictable user-selected length.
.Ss First barrier: the location of the \&"lock-sector".
During initialization, up to four independent but mutually aware
.Dq lock
sectors are written to the device in randomly chosen
locations.
These lock-sectors contain the 2048 random bit master-key and a number
of parameters of the layout geometry (more on this later).
Since the entire device will contain isotropic data, there is no
short-cut to rapidly determine which sequence of bytes contain a lock-sector.
.Pp
To locate a lock-sector, a small piece of data called the
.Dq metadata
and the key-material must be available.
The key-material decrypts the
metadata, which contains the byte offset on the device where the
corresponding lock-sector is located.
If the metadata is lost or unavailable but the key-material is at
hand, it would be feasible to do a brute force scan where each byte offset
of the device is checked to see if it contains the lock-sector data.
.Ss Second barrier: decryption of the master-key using key-material.
The lock-sector contains an encrypted copy of an architecture neutral
byte-sequence which encodes the fields of the lock-structure.
The order in which these fields are encoded is determined from the key-material.
The encoded bytestream is encrypted with 256bit AES in CBC mode.
.Ss Third barrier: decryption of the sector key.
For each sector, an MD5 hash over a
.Dq salt
from the lock-sector and the sector number is used to
.Dq cherry-pick
a subset of the master key,
which hashed together with the sector offset through MD5 produces the
.Dq kkey ,
the key which encrypts the sector key.
.Ss Fourth barrier: decryption of the sector data.
The actual payload of the sector is encrypted with 128 bit AES in CBC mode
using a single-use random bits key.
.Ss Examining the reverse path
Assuming an attacker knows an amount of plaintext and has managed to
locate the corresponding encrypted sectors on the device, gaining access
to the plaintext context of other sectors is a daunting task:
.Pp
First he will have to derive from the encrypted sector and the known plain
text the sector key(s) used.
At the time of writing, it has been speculated that it could maybe be
possible to break open AES in only 2^80 operations; even so, that is still
a very impossible task.
.Pp
Armed with one or more sector keys, our patient attacker will then go
through essentially the same exercise, using the sector key and the
encrypted sector key to find the key used to encrypt the sectorkey.
.Pp
Armed with one or more of these
.Dq kkeys ,
our attacker has to
run them backwards through MD5.
Even though he knows that the input to MD5 was 24 bytes and has the value
of 8 of these bytes from the sector number, he is still faced with 2^128
equally likely possibilities.
.Pp
Having successfully done that, our attacker has successfully discovered
up to 16 bytes of the master-key, but is still unaware which 16 bytes,
and in which other sectors any of these known bytes contribute to the kkey.
.Pp
To unravel the last bit, the attacker has to guess the 16 byte random-bits
salt stored in the lock-sector to recover the indexes into the masterkey.
.Pp
Any attacker with access to the necessary machine power to even attempt
this attack will be better off attempting to brute-force the pass-phrase.
.Ss Positive denial facilities
Considering the infeasibility of the above attack,
gaining access to the pass-phrase will be of paramount importance for an
attacker,
and a number of scenarios can be imagined where undue pressure will be
applied to an individual to divulge the pass-phrase.
.Pp
A
.Dq Blackening
feature provides a way for the user, given a moment of
opportunity, to destroy the master-key in such a way that the pass-phrase
will be acknowledged as good but access to the data will still be
denied.
.Ss A practical analogy
For persons who think cryptography is only slightly more interesting than
watching silicon sublimate the author humbly offers this analogy to the
keying scheme for a protected device:
.Pp
Imagine an installation with a vault with walls of several hundred meters
thick solid steel.
This vault can only be feasibly accessed using the
single key, which has a complexity comparable to a number with 600 digits.
.Pp
This key exists in four copies, each of which is stored in one of
four small safes, each of which can be opened
with unique key which has a complexity comparable to an 80 digit
number.
.Pp
In addition to the masterkey, each of the four safes also contains
the exact locations of all four key-safes which are located in
randomly chosen places on the outside surface of the vault where they
are practically impossible to detect when they are closed.
.Pp
Finally, each safe contains four switches which are wired to a bar
of dynamite inside each of the four safes.
.Pp
In addition to this, a keyholder after opening his key-safe is
also able to install a copy of the master-key and re-key any of
key-safes (including his own).
.Pp
In normal use, the user will open the safe for which he has the key,
take out the master-key and access the vault.
When done, he will lock up the master-key in the safe again.
.Pp
If a keyholder-X for some reason distrusts keyholder-Y, she
has the option of opening her own safe, flipping one of the switches
and detonating the bar of dynamite in safe-Y.
This will obliterate the master-key in that safe and thereby deny
keyholder-Y access to the vault.
.Pp
Should the facility come under attack, any of the keyholders can detonate
all four bars of dynamite and thereby make sure that access to the
vault is denied to everybody, keyholders and attackers alike.
Should the facility fall to the enemy, and a keyholder be forced to apply
his personal key, he can do so in confidence that the contents of his safe
will not yield access to the vault, and the enemy will hopefully realize
that applying further pressure on the personnel will not give access to
the vault.
.Pp
The final point to make here is that it is perfectly possible to
make a detached copy of any one of these keys, including the master
key, and deposit or hide it as one sees fit.
.Ss Steganography support
When the device is initialized, it is possible to restrict the encrypted
data to a single contiguous area of the device.
If configured with care, this area could masquerade as some sort of
valid data or as random trash left behind by the systems operation.
.Pp
This can be used to offer a plausible deniability of existence, where
it will be impossible to prove that this specific area of the device
is in fact used to store encrypted data and not just random junk.
.Pp
The main obstacle in this is that the output from any encryption algorithm
worth its salt is so totally random looking that it stands out like a sore
thumb amongst practically any other sort of data which contains at least
some kind of structure or identifying byte sequences.
.Pp
Certain file formats like ELF contain multiple distinct sections, and it
would be possible to locate things just right in such a way that a device
contains a partition with a filesystem with a large executable,
.Pq Dq "a backup copy of my kernel"
where a non-loaded ELF section is laid out
consecutively on the device and thereby could be used to contain a
.Nm
encrypted device.
.Pp
Apart from the ability to instruct
.Nm
which those sectors are, no support is provided for creating such a setup.
.Ss Deployment suggestions
For personal use, it may be wise to make a backup copy of the masterkey
or use one of the four keys as a backup.
Fitting protection of this key is up to yourself, your local circumstances and
your imagination.
.Pp
For company or institutional use, it is strongly advised to make a copy
of the master-key and put it under whatever protection you have at your
means.
If you fail to do this, a disgruntled employee can deny you access to
the data
.Dq "by accident" .
(The employee can still intentionally deny access by applying another
encryption scheme to the data, but that problem has no technical solution.)
.Ss Cryptographic strength
This section lists the specific components which contribute to the cryptographic
strength of
.Nm .
.Pp
The payload is encrypted with AES in CBC mode using a 128 bit random
single-use key
.Pq Dq "the skey" .
AES is well documented.
.Pp
No IV is used in the encryption of the sectors, the assumption being
that since the key is random bits and single-use, an IV adds nothing to the
security of AES.
.Pp
The random key is produced with
.Xr arc4rand 9
which is believed to do a respectable job at producing unpredictable bytes.
.Pp
The skey is stored on the device in a location which can be derived from
the location of the encrypted payload data.
The stored copy is encrypted with AES in CBC mode using a 128 bit key
.Pq Dq "the kkey"
derived
from a subset of the master key chosen by the output of an MD5 hash
over a 16 byte random bit static salt and the sector offset.
Up to 6.25% of the masterkey (16 bytes out of 2048 bits) will be selected
and hashed through MD5 with the sector offset to generate the kkey.
.Pp
Up to four copies of the master-key and associated geometry information
is stored on the device in static randomly chosen sectors.
The exact location inside the sector is randomly chosen.
The order in which the fields are encoded depends on the key-material.
The encoded byte-stream is encrypted with AES in CBC mode using 256 bit
key-material.
.Pp
The key-material is derived from the user-entered pass-phrase using
512 bit SHA2.
.Pp
No chain is stronger than its weakest link, which usually is poor pass-phrases.
.Sh SEE ALSO
.Xr gbde 8
.Rs
.%A Poul-Henning Kamp
.%T "Making sure data is lost: Spook-strength encryption of on-disk data"
.%R "Refereed paper, NORDU2003 conference"
.Re
.Sh HISTORY
This software was developed for the
.Fx
Project by
.An Poul-Henning Kamp
and NAI Labs, the Security Research Division of Network Associates, Inc.\&
under DARPA/SPAWAR contract N66001-01-C-8035
.Pq Dq CBOSS ,
as part of the
DARPA CHATS research program.
.Sh AUTHORS
.An "Poul-Henning Kamp" Aq phk@FreeBSD.org
|