Using PGP/PEM encryption

This document is formatted roughly like a FAQ. Here are the section headings:

What is PGP/PEM encryption?

 PGP and PEM are programs to allow you and a second party to communicate in a way which does not allow third parties to read them, and which certify that the person who sent the message is really who they claim they are.

PGP and PEM both use RSA encryption. The U.S. government has strict export controls over foreign use of this technology, so people outside the U.S. may have a difficult time finding programs which perform the encryption.

How are they used in HTTP?

 We have implemented a system by which NCSA Mosaic and NCSA httpd call external programs which encrypt and decrypt their communications and thus provide secure communications between the server and the client and ensure that a user is who he/she says they are. This system currently has hooks for PEM encryption as well as PGP encryption. As interest in this area grows, more will most likely be added.

What do I need to use it?

 You will need a working copy of either Pretty Good Privacy or RIPEM to begin with. You should be familiar with the program and have generated your own public/private key pair.

You should be able to use the TIS/PEM program with the PEM authorization type. I haven't tried it. This tutorial is written assuming that you are using RIPEM.

How do the keys get distributed?

 Currently, we have implemented this protocol with PEM and PGP using local key files on the server side, and on the client side with PEM using finger to retrieve the server's public key.

As you can tell, parties who wish to use Mosaic and httpd with PEM or PGP encryption will need to communicate beforehand and find a tamper-proof way to exchange their public keys.

How bulletproof is it?

 Pioneers get shot full of arrows. This work is currently in the experimental stages and thus may have some problems that I have overlooked. There aren't any that I can see but I've been looking at it for a long time. There may be some quirks in the additions to Mosaic and httpd as well. In particular, error recovery is not always as helpful as it could be.

The only known problem that I know about is that the messages are currently not timestamped. This means that a malicious user could record your encrypted message with a packet sniffer and repeat it back to the server ad nauseum. Although they would not be able to read the reply, if the request was something you were being charged for, you may have a large bill to pay by the time they're through.

Installation: The Server Side

 First, you must compile httpd with CFLAGS set to -DPEM_AUTH. This will enable the PEM and PGP authentication directives.

Next, look in the support/auth directory. This directory contains your encryption and decryption scripts, as well as bins for your remote users' public keys.

Edit ripem-dec, ripem-enc, pgp-enc, and pgp-dec and follow the instructions therein. You need not set up PGP if you don't plan to use it, and same with RIPEM.

Now, edit your server configuration file, usually conf/httpd.conf. You will want to add three new directives for PGP and three new directives for PEM depending on which you plan to use (or both). 

PEMEncryptCmd /usr/local/etc/httpd/auth/ripem-enc
PGPEncryptCmd /usr/local/etc/httpd/auth/pgp-enc
These directives change what executables httpd will look for when it is trying to encrypt its reply to a client. Edit the pathnames to taste.

PEMDecryptCmd /usr/local/etc/httpd/auth/ripem-dec
PGPDecryptCmd /usr/local/etc/httpd/auth/pgp-dec
These directives change what executables httpd will look for when it is trying to decrypt the client's request. 

PEMServerEntity webmaster@foobar.org
PGPServerEntity webmaster@foobar.org
These directives set your entity name. This should be the same as the name you place on the public/private keys you generate for your server. If you make these directives different than the key names, your server and its client will become hopelessly confused.

You are now ready to protect directories of your server with this authorization scheme.

For a directory you want to protect, you should first set its AuthType. Use AuthType PGP for a directory you are protecting with PGP and AuthType PEM for a directory you are protecting with PEM.

The require directive accepts key names as its arguments. The AuthGroupFile directive is valid as well, to create groups of keys. A full example: 

AuthType PEM
AuthGroupFile /httpd/.htgroup-pem
Options None
<Limit GET>
require user robm@ncsa.uiuc.edu
require group pemusers
Let's say /httpd/.htgroup-pem reads:

pemusers: pls@ncsa.uiuc.edu In this case, this directory will be protected with PEM encryption and will require that only users robm@ncsa.uiuc.edu and pls@ncsa.uiuc.edu be allowed to access that directory.

Installation: the Client Side

 First, get a copy of Mosaic/X 2.2. If it hasn't been released yet, be patient. Compile it with -DPEM_AUTH to enable PEM/PGP authentication.

Follow the instructions in each of the scripts in the auth subdirectory to customize them to your setup.

There are six new X resources which have been defined for PEM/PGP authentication. They are: 

Mosaic*pemEncrypt: /X11/robm/Mosaic/auth/ripem-enc
Mosaic*pemDecrypt: /X11/robm/Mosaic/auth/ripem-dec
Mosaic*pemEntity: robm@ncsa.uiuc.edu
Mosaic*pgpEncrypt: /X11/robm/Mosaic/auth/pgp-enc
Mosaic*pgpDecrypt: /X11/robm/Mosaic/auth/pgp-dec
Mosaic*pgpEntity: robm@ncsa.uiuc.edu
You should change the Encrypt and Decrypt entries to reflect where you are going to install your encryption and decryption crypts. You should change the Entity lines to the key name you have given the server maintainers for yourself. If you don't, bad things will happen.

What does the protocol look like?

 This protocol is almost word-for-word a copy of Tony Sander's RIPEM based scheme, generalized a little. Below, wherever you see PEM you can replace it with PGP and get the same thing.

Client: 

GET /docs/protected.html HTTP/1.0
UserAgent: Mosaic/X 2.2
Server: 

HTTP/1.0 401 Unauthorized
WWW-Authenticate: PEM entity="webmaster@hoohoo.ncsa.uiuc.edu"
Server: NCSA/1.1
Client: 

GET / HTTP/1.0
Authorization: PEM entity="robm@ncsa.uiuc.edu"
Content-type: application/x-www-pem-request

--- BEGIN PRIVACY-ENHANCED MESSAGE ---
this is the real request, encrypted
--- END PRIVACY-ENHANCED MESSAGE ---
Server: 

HTTP/1.0 200 OK
Content-type: application/x-www-pem-reply

--- BEGIN PRIVACY-ENHANCED MESSAGE ---
this is the real reply, encrypted
--- END PRIVACY-ENHANCED MESSAGE ---
That's it.

Almost all of this stuff is my fault (including the implementations), so direct comments about it to me.

Rob McCool, robm@ncsa.uiuc.edu