I never saw the posting below here, only on sci.crypt. It seems relevant to the present discussion. I did take a look at the source code to stealth pgp; it is quite simple and nicely done. However it doesn't do anything special about disguising the encrypted session key. It just strips off the two length bytes PGP puts at the front. Not only is this revealing, it also doesn't always work. Apparently the program can't always reconstruct the original mpi length if the encrypted key's size happens to be much less than the modulus. It does print a warning in that case that the file won't be able to be de-stealth'd successfully. Stealth PGP is a nice start but it needs to be improved to be truly stealthy.
From: qwerty@netcom.com (-=Xenon=-) Subject: Stealth for PGP Available. Message-ID: <qwertyCLpztC.4n8@netcom.com> Organization: PGP Info Clearinghouse. Date: Thu, 24 Feb 1994 08:34:24 GMT Lines: 188
-----BEGIN PGP SIGNED MESSAGE-----
I have placed Stealth1.0 up for anonymous ftp at netcom.com in /pub/qwerty as stealth.tar.Z. I am not the author of Stealth. "Henry Hastur" is the author. I will gladly forward mail to him if you tell me to do so. My address is qwerty@netcom.com. So far it is only written for Unix, so don't e-mail me asking for a DOS version or whatever. I'm just a cheerleader and propagandist, who finally got some very kind soul to "write me a real encryptor".
Why Stealth? Wired magazine just gave a plug for the Mac program "Stego" (ftp to sumex-aim.stanford.edu in /info-mac/cmp) by Romana Machado, which will, like any steganograph, hide a message in a carrier message. Problem is, anybody can reverse it, and if you are using PGP, immediately tell you are one of those drug dealing, child molesting, cryptography users ;-). Once Stealth is incorporated into such steganographs, then people can't easily tell that what they've extracted out of that picture of Madonna is even a message at all. If they don't have the right secret key to decrypt it, then all they get is garbage. "What message?!"
Questions:
1) What does Stealth have to do with farming? 2) How do stealth encryptors make the Clipper chip look like a stupid waste of YOUR tax money?
[Answers: 1) Zip, 2) By demonstrating that even if the Clipper keys are stolen, without your secret key they can't even tell you are using PGP].
-=Xenon=-
Stealth V1.0 by Henry Hastur ----------------------------
Stealth is a simple filter for PGP which strips off all identifying header information to leave only the encrypted data in a format suitable for steganographic use. That is, the data can be hidden in images, audio files, text files, CAD files, and/or any other file type that may contain random data, then sent to another person who can retrieve the data from the file, attach headers, and PGP decrypt it.
Stealth is not intended to replace the standardized methods of using encryption (e.g. ASCII-armoured PGP email) ; in an ideal world we would all be able to send openly encrypted mail or files to each other with no fear of reprisals, however there are often cases when this is not possible, either because the local government does not approve of encrypted communication, or perhaps because you are working for a company that does not allow encrypted email but doesn't care about Mandelbrot GIFs. This is where Stealth and steganography can come into play.
Compiling - - ---------
Stealth has currently only been tested on BSD and SVR4 Unix (and as such should work with most varieties of Unix), with both non-ANSI compilers and ANSI compilers with 'minimal ANSI' flags. In order to compile the program, you should just be able to extract the files from the tar file provided, then type 'make'. If that fails you may need to change the definition of CC and CFLAGS in the makefile to get it to compile. On machines with gcc, the GNU C compiler, Stealth can be compiled by simply changing the 'CC=cc' line in makefile to 'CC=gcc'.
Stealth has not yet been tested on MS-DOS, but the only likely problems are with 16-bit integers (you may need to change some occurrences of int to long in order to get stealth to work), and you will need to remove the -DUNIX flag from compilations. Hopefully, version 1.1 will be released shortly with full DOS compatibility.
Usage - - -----
Stealth always reads from its standard input and writes to the standard output, though when adding headers to data the data has to be stored in a temporary file (see Security Concerns below).
Command line arguments :
-c Conventional encryption used rather than public key -a Add headers (defaults to strip headers) -v Verbose output.
Stealth needs to be able to find your pubring.pgp file, which it does by first checking in the directory pointed to by $PGPPATH, then the current directory.
Examples - - --------
To encrypt a file with PGP and store it in the file pgp.stl prior to sending :
pgp -ef < secrets.dat | stealth > pgp.stl
To encrypt a file with conventional (IDEA) encryption, and pass to a steganography program called steg_program :
pgp -fec < secrets.dat | stealth -c | steg_program
To take the output from a steganographic extraction tool, add headers for key "Your Id", and decrypt :
steg_program | stealth -a "Your Id" | pgp -f > secrets.dat
To take the conventionally encrypted output from a steg program, attach headers and decrypt :
steg_program | stealth -ac | pgp -f > secrets.dat
Limitations - - -----------
Files can be signed, but can only be encrypted to one recipient - extra RSA headers for all but the first recipient will be stripped from the file. In addition, if you specify conventional encryption but pass an RSA-encrypted file into the filter the RSA-block will be stripped. In either case, stealth will print out warnings to inform you of this.
Stealth provides no support for ASCII-armoured PGP messages - it will only work with the binary output format, and the output will have to be converted to a useable form after processing, either with a steganography program or a standard utility such as uuencode.
Finally, for technical reasons there are potential problems with public keys of size (typically) 2^n + 1 or 2^n + 2 (e.g. 513 or 1026). If you are encrypting to a key of a peculiar size, it's possible that the algorithm used to add headers could fail, but fortunately this can be detected while stripping the headers, and a warning will be printed. If this warning appears, you will probably want to encrypt the data again until a suitably sized RSA-block is created.
It is NOT neccesary to remove garbage data that the steganography program may have added to the end of the PGP-encrypted data. PGP output contains an encrypted end-of-file mark that allows the program to decrypt correctly and ignore any trailing garbage.
Security Concerns - - -----------------
After passing through the stealth filter, the PGP-encrypted data is essentially white noise, with no identifying marks, and whilst it may well have enough peculiarities for an expert cryptanalyst to recognize it as encrypted data, the probability is much less than would be the case with a PGP header identifying the recipient attached.
One other concern is that stealth has to create a temporary file when reading in data to attach headers, and depending on the build options chosen the program will store it in either $PGPPATH, the current directory or /tmp. On Unix machines, the file will be deleted as soon as it is opened, making it difficult to capture, but on other operating systems the file will only be deleted when it has been used. (In either case the file will be zeroed before being closed).
In addition, some operating systems will use temporary files on your disk to emulate unix pipes (e.g. MS-DOS) - these files will not be zeroed when finished with !
Export Restrictions - - -------------------
Stealth is probably not covered by current export restrictions under the US ITAR regs, but I'm not a lawyer, so if in doubt check it out yourself. It was written outside the US and imported, so should soon be available on some European ftp sites as well as US sites.
Henry Hastur
-----BEGIN PGP SIGNATURE----- Version: 2.3
iQCVAgUBLWwfawSzG6zrQn1RAQEjlQP/d85opZ6aMXfFpDu/KQz0DAzFxXNjiSZy 9l56kIUkmAx0aT5qF3UAg2UTDagiJlFNz6UXFhEKBkDV4JLJPq5C8HtsdpCiSLWr vVD2IEdASEd9ALCQMLS/YxVv6GG85n+phyoEV5ALsD1f0y4Nbk2Gfb+rexk5rN3a hFYRUekVuNY= =pkdK -----END PGP SIGNATURE-----
However it doesn't do anything special about disguising the encrypted session key.
And as Hal and I have been discussing, that's not at all an obvious problem. A filter for PGP messages cannot make them completely random for all the reasons presented. The session keys must be generated differently if the encrypted form if them is to have a flat distribution. To wit, PGP itself must change in order to make a random PGP output format. Eric
participants (2)
-
Hal -
hughes@ah.com