What Is This?
Try It Out!
Building the Server
There are three parts to the RPOW system: client, host and server. The source tarball available here includes all three. Each part is of different utility for different readers.
The client code is most generally useful. It builds a library which communicates with the RPOW server and maintains a local file of available RPOW tokens. There is also a command-line front end program which is intended as a demo of how to use the library. The code is released with an open source license and I encourage readers to add other front ends, such as a GUI, to make the system more functional and useful.
The server code is the core of the security of the RPOW system. The unique claim of RPOW is that anyone can build an exact copy of the program running on the RPOW server using this code, and remotely verify that this is in fact the software they are communicating with. See the page on Building the Server for extensive instructions on building the RPOW server memory image.
The host code is of less general interest. This is the component that runs on the server where the RPOW secure coprocessor card is installed. Its main function is to listen for net connections and pass them to the card, returning the results. It also supports housekeeping functions for initialization, key rollover, etc. The host code will mostly be of interest to those who are thinking of acquiring an IBM 4758 card and running their own RPOW server. I will be happy to assist anyone who wants to undertake such a project.
Here is the tarball with the source code for the RPOW project:
sha1sum: c16fbf594e64272879ade6892f00a743f6e05869 rpow-1.2.0.tar.gz
Building the client
Building the client requires two libraries. One is hashcash, available from www.hashcash.org. You need version 1.16 or later. However, there is a bug in version 1.16 which causes occasional errors on generating large hashcash tokens. Until this is fixed, I am making available a patched version as hashcash-1.16-patched.tgz.
RPOW also depends on the OpenSSL library being available. This comes installed on all modern Linux systems, but on Windows you will have to download it and build it there.
When building hashcash, follow the instructions in the Makefile to use OpenSSL for the token generation. This is necessary for the hashcash library to be compatible with the RPOW code.
On Linux systems, go into the client subdirectory and type make. This will create the executable rpowcli.
The RPOW client Makefile assumes that the hashcash source is downloaded in a directory called hashcash which is a sibling of the main rpow directory. If you put your hashcash source elsewhere or call it something else than hashcash, create a symbolic link as a sibling of rpow called hashcash which points at your hashcash directory; or alter the RPOW client Makefile to point at your hashcash source.
For Windows I recomend the use of the MinGW package, which allows building native Windows executables. You can build OpenSSL and hashcash using this package, and then build rpowcli to link to these.
This is the pre-built client command line for Linux systems:
This is the pre-built client command line for Windows systems:
And this is the pre-built client command line for Mac OS X systems:
Running the Client
The provided RPOW client is a command line based program intended to demonstrate the functionality of the system. The program is split into a command line driver program and a library, where the driver makes calls into the library to perform the needed functions. It is hoped that future development will result in more functional and attractive front ends which use the same library interface.
rpowcli maintains several files which on Linux go by default into a directory named ~/.rpow (where ~ is the user's home directory). Users should create that directory before running the client. Windows users should set an environment variable named RPOW_PATH to be the directory where rpowcli will store its files.
In that directory the user should create a file called config which will contain directives read by the RPOW client. The only directive supported at present is host, which tells where to look for the RPOW server. It should be set as follows, initially:
host = rpow.net:4902
Once the ~/.rpow/config file (or corresponding one on Windows) is set up with this entry, you are ready to try the client.
(Note that RPOW now provides support for using the Tor anonymizing system for access to the RPOW server. Users who want to take extra precautions to guard their privacy may wish to set up Tor and enable RPOW to use it even before their first access to the RPOW server. This way, the RPOW server will never see the Internet address where the requests are coming from. For details, see the Privacy page.)
rpowcli takes a first argument which is a command name, followed by arguments. The various commands are as follows:
- This should be the first command run. It contacts the server identified in the host command in the config file, and downloads a certificate chain from that server which includes the keys for communicating with the server and for signing RPOWs. It then passes this certificate chain through a stringent validation test to be sure that it is in fact an RPOW server. See the security page for more details of how the client validates the server. If the server passes the tests, the client extracts the server keys from the certificate chain and saves them in the .rpow directory. It also prints the name of the hashcash "resource string" that should be used with this particular RPOW server in case you want to use the hashcash program to generate (or "mint") values which will be acceptable by the RPOW server. This is a long string composed of three hexadecimal numbers separated by dashes, followed by ".rpow.net". Note that getkeys is intended to be used only with a fresh install and will wipe out any saved RPOW tokens in the .rpow directory.
- This works the same as the getkeys command except that it retains any saved RPOW tokens. It can be given at any time if you want to make sure you have valid keys for the server. If the server has gone through a key rollover (described in the World of RPOW page), clients must use the rekey command to acquire the new keys from the server. Their old tokens will still be good, for a period of time, although they should be exchanged when convenient. This command can also be used to print out the hashcash resource string in case you want to use the hashcash program to create POW tokens as described below.
- gen n
- This command generates an RPOW token worth n bits. n must be in the range from 20 to 50. This command is one of the main ways that you can receive RPOW tokens. It uses the hashcash library to generate a version 1 hashcash string worth n bits, then exchanges that POW token at the RPOW server for an equal-value RPOW token. See the theory page for how POW and RPOW tokens relate to each other. The received RPOW token is stored in a file in the .rpow directory named rpows.dat.
- This command runs continually and generates RPOW tokens until it is interrupted. It generates eight POW tokens and then exchanges them at the server for an RPOW token. The size of the tokens is adjusted dynamically so that it will take from ten minutes to one hour to generate the eight tokens. If you interrupt the program with control-C before it has generated eight tokens, they will not be exchanged and the last few minutes of calculation will be lost. That's not usually something to worry about; this command is designed to be run overnight or when the computer is not otherwise being used, to help build up a supply of RPOWs.
- Count how many RPOW tokens you have of different values in the rpows.dat file and display the counts.
- out n
- Find an output token worth n bits in the rpows.dat file and output it to standard out, in base64 (text) format. Its intended use is for sending RPOWs to other people. If you don't have a token of n bits, but you do have larger ones, this command will perform exchanges at the server to break the larger RPOWs into smaller sizes so that it can output the requested size.
- Read an RPOW or POW token from standard in, in text format, and exchange it at the server for a new RPOW token, saving that in a file. This prints a message telling the value of the token received. It is intended to be used to receive RPOW tokens from other people. Exchanging the token at the server is the mechanism by which the RPOW client verifies that the incoming token is legal: properly formatted, and not used before. This command can also be used to receive POW tokens created by the hashcash program. Run it as:hashcash -m -q -b bits -r rpow-resource | rpowcli inGet the correct rpow-resource from the string printed by the rekey command. This value will not change for the life of the server. This command is equivalent to doing:rpowcli gen bitsbut the hashcash program may have optimizations which have not made it into the RPOW client, so it might be faster.
- exchange inval ... 0 outval ...
- This exchanges RPOW tokens from your rpows.dat file for new RPOWs of equal value. List the values of the RPOW tokens you want to exchange, one after another on the command line. Then add a zero, and then list the values of the new RPOW tokens you want to create. The total value of the new RPOW tokens must equal the total value of the old ones, keeping in mind the rule that each RPOW token is worth twice the previous one. So you can, for example, exchange two RPOW tokens worth 20 for one worth 21; or exchange one worth 30 for three worth 28 and two worth 27. There is a limit in exchange operations of no more than eight input and/or output tokens per exchange. If you have more tokens than that you will have to do multiple exchange operations. This command is meant to be used to split or combine RPOW tokens as described in the theory page. It can also be used to replace all of your saved RPOW tokens when there has been a key rollover, as described in World of RPOW.
- This combines smaller RPOW tokens from your rpows.dat file into larger ones by doing exchanges at the server. The result will be that you will only have one RPOW token of any given size (since two of the same size can be combined to make one of the next larger size), up to the maximum size of 50. You could have more than one size-50 RPOW token but that would take a long time to generate (perhaps 50 year-2000-computer-years!). Note that the out command will break up RPOWs into smaller units if necessary.
- This receives and prints out an extensive block of information from the server with information about the IBM 4758 and the software running on it. Most of it is pretty technical and will not be of much interest unless you know a lot about the 4758. It does include some information about free memory on the system, and it lists how many keys the server presently knows about and trusts. (See the World of RPOW page for information about how servers come to trust multiple RPOW signing keys.)
Those are all of the rpowcli commands. Here is a typical usage example:$ rpowcli rekey Retrieving certificate chain from server... Validating certificate chain... IBM root key signature on class key validated Class key signature on device key validated Device key signature on miniboot key validated Miniboot key signature on OS key validated OS key signature on application key validated Cert chain validates OK, checking hashes... Hash is acceptable! Proceed... Successfully read keys for 71672bc400000017-000003c2-0001.rpow.net $ rpowcli gen 20 $ rpowcli gen 20 $ rpowcli count 2 rpows in rpow data file /home/user/.rpow/rpows.dat: value 20: 2 $ rpowcli exchange 20 20 0 21 $ rpowcli count 1 rpows in rpow data file /home/user/.rpow/rpows.dat: value 21: 1 $ rpowcli exchange 21 0 20 20 $ rpowcli count 2 rpows in rpow data file /home/user/.rpow/rpows.dat: value 20: 2 $ rpowcli out 20 > rp20 $ rpowcli count 1 rpows in rpow data file /home/user/.rpow/rpows.dat: value 20: 1 $ cat rp20 AQAAABSKLQ6BRhHTEyPNUhFHOS5Y289wmlnSfdxFI6H7Ydosoj/kogkG+pc9cWcrxAAAABcAAAOpAAEA AACAu5XQT24VcgiL/0vXIUG99JGK2lpoIS2ZFoL+9dxpTOXEfS8vqN31PRLgrKHExfDeDOxZPY44JsAN yYfjQHGw+63j5Fvcb3hNydvMg8A7EHNBqCm1OYiejoUpZXOxfolVGR0g0tDkR8w/E6hL6J7oUgODckug odFcFRzChlVTUKU= $ rpowcli in < rp20 Received rpow item of value 20 $ rpowcli count 2 rpows in rpow data file /home/user/.rpow/rpows.dat: value 20: 2 $ rpowcli out 20 | rpowcli in Received rpow item of value 20 $ rpowcli count 2 rpows in rpow data file /home/user/.rpow/rpows.dat: value 20: 2 $ rpowcli consolidate $ rpowcli count 1 rpows in rpow data file /home/user/.rpow/rpows.dat: value 21: 1
At this point the RPOW client is relatively immature and has not been built on many hosts. It is not a particularly complicated program, though, so hopefully it will be possible to easily fix up any portability problems. Please notify the author if you need to make any changes to get it to build on your host system.
If the server fails to respond, it may be down temporarily for maintenance, although at this point I am to provide a high degree of availability for the server. It is also possible that the server may become overloaded if many people try out the RPOW client at once. I aim to acquire additional IBM 4758 boards to experiment with, and I can make those available as RPOW servers if necessary.