.RI [ item ...]
.br
.B encrypt
-.RB [ \-a ]
+.RB [ \-aC ]
.RB [ \-k
.IR tag ]
.RB [ \-f
.RI [ file ]
.br
.B decrypt
-.RB [ \-aqv ]
+.RB [ \-aqvC ]
.RB [ \-f
.IR format ]
.RB [ \-o
.TP
.B ec
This is the elliptic-curve analogue of
-.BR dh . Use the
+.BR dh .
+Use the
.B ec
algorithm of the
.BR key (1))
attribute then the
.I hash
in the
-.I kemalgspec is used; if that is absent then the default of
+.I kemalgspec
+is used; if that is absent then the default of
.B rmd160
is used. Run
.B catcrypt show hash
Write output to
.I file
rather than to standard output.
+.TP
+.B "\-C, \-\-nocheck"
+Don't check the public key for validity. This makes encryption go much
+faster, but at the risk of using a duff key.
.SS decrypt
The
.B decrypt
instead of to standard output. The file is written in binary mode.
Fixing line-end conventions is your problem; there are lots of good
tools for dealing with it.
+.TP
+.B "\-C, \-\-nocheck"
+Don't check the private key for validity. This makes decryption go much
+faster, but at the risk of using a duff key, and possibly leaking
+information about the private key.
.PP
Output is written to standard output in a machine-readable format.
Major problems cause the program to write a diagnostic to standard error
.TP
.B "DATA"
The plaintext follows, starting just after the next newline character or
-sequence. This is only produced if main output is being sent to
+sequence. This is only produced if main output is also being sent to
standard output.
.TP
.BI "INFO " note
All output written has been checked for authenticity. However, output
can fail madway through for many reasons, and the resulting message may
therefore be truncated. Don't rely on the output being complete until
-.B OK is printed or
+.B OK
+is printed or
.B catcrypt decrypt
exits successfully.
.SS "encode"
warranties. But it does avoid the usual problem with separate signing
and encryption that a careful leak by the recipient can produce evidence
that you signed some incriminating message.
+.PP
+Note that
+.BR catcrypt 's
+signatures do
+.I not
+provide `non-repudiation' in any useful way. This is deliberate: the
+purpose of signing is to convince the recipient of the sender's
+identity, rather than to allow the recipient to persuade anyone else.
+Indeed, given an encrypted and signed message, the recipient can
+straightforwardly construct a new message, apparently from the same
+sender, and whose signature still verifies, but with arbitrarily chosen
+content.
.SH "CRYPTOGRAPHIC THEORY"
Encryption of a message proceeds as follows.
.hP 0.
scheme; use the next bits to key a message authentication code.
.hP 4.
If we're signing the message then extract 1024 bytes from the keystream,
-sign them, and emit a packet containing the signature. The signature
-packet doesn't contain the signed message, just the signature.
+sign the header and public value, and the keystream bytes; emit a packet
+containing the signature. The signature packet doesn't contain the
+signed message, just the signature.
.hP 5.
Split the message into blocks. For each block, pick a random IV from
the keystream, encrypt the block and emit a packet containing the
.BR hashsum (1),
.BR keyring (5).
.SH AUTHOR
-Mark Wooding, <mdw@nsict.org>
+Mark Wooding, <mdw@distorted.org.uk>