NAME

hashcash - hashcash anti-spam / denial of service counter-measure tool


SYNOPSIS

Mint (create) token:

hashcash [ options ] [ -b bits ] [ resource ]>

Check token:

hashcash -c [ options ] -b bits -r resource [ token ]

Speed estimates:

hashcash -s [ options ] [ -b bits ]

Purge database:

hashcash -p now [ -j resource ] [ -k ] [ -t time ] [ -u ]

Count collision bits:

hashcash -w [ options ] [ token ]

Get resource name:

hashcash -n [ options ] [ token ]

Get time remaining until expiry

hashcash -l [ options ] -e period [ token ]


DESCRIPTION

Hashcash is a denial-of-service counter measure tool. It's main current use is to help hashcash users avoid losing email due to content based and blacklist based anti-spam systems.

The hashcash tool allows you to create hashcash tokens to attach to emails you send, and to verify hashcash tokens attached to emails you receive. Email senders attach hashcash tokens with the X-Hashcash: header. Vendors and authors of anti-spam tools are encouraged to exempt mail sent with hashcash from their blacklists and content based filtering rules.

A hashcash token constitutes a proof-of-work which takes a parameterizable amount of work to compute for the sender. The recipient can verify received tokens efficiently. This package also includes a sha1 implementation which behaves somewhat like md5sum, but with SHA1.

The hashcash proof-of-work function is to compute partial hash collisions (the SHA1 hash function is used). By choosing the number of bits of collision, the work required to compute a collision can be made arbitrarily expensive -- from fractions of a second to minutes or hours. The verification is efficient requiring the same small cost whatever the collision size.

For more detailed discussion of other applications hashcash has been used for see http://www.hashcash.org/


USAGE NOTES

In this man page a resource name is the name of the service or address the token is created for. In the case of email, the resource name is the recipient's email address in the form user@domain.com.

Minting tokens

If neither the -c or -s option are given, it is assumed that you want to mint a token.

The resource name (recipient's email address) to mint the token against can be passed as an argument, or if omitted is read from stdin. If stdin is a tty the user is prompted, if stdin is a pipe the resource name is just silently read. The desired collision size can be specified with the -b option. If no collision size is specified, the default is 20 bits. As a convenience, if stdin is a tty and no collision size is given with the -b flag, the user is prompted for a collision size.

Checking tokens

The -c flag must be given to check tokens. The token to check can be given as an argument to hashcash. If no token is given the token is read from stdin. If stdin is a tty the user will be prompted, if stdin is a pipe the token is just silently read. A resource name (the recipient's email address) can be given with the -r option. If a resource name is given the resource name is compared to the resource name in the token, if they do not match, the token is rejected.

Note: if no resource name is given the token is anyway checked to see if it is otherwise valid, but it could be minted for a different resource, which would allow tokens to be reused across different resources, so hashcash will return unchecked exit code on exit.

Tokens are by default considered to be valid forever. The validity period can be changed using the -e flag.

If the token has expired or has a date in the future the token is rejected and the program exits immediately.

If a required collision size is given with the -b flag, the tokens value is computed and compared, if the token has insufficent value it is rejected, and the program exits immediately. If the -b flag is not given, the token could be of any size, so hashcash will return unchecked exit code on exit.

If the token is double spent the token is rejected. Double spending protection is discussed in more detail below in Double Spending Protection. If double spending protection is not enabled, the token could be double spent, so hashcash will return unchecked exit code (exit code 2) on exit.

The -w flag can be used to request that the number of bits of the collision are counted and displayed; exit code unchecked (exit code 2) is returned on exit. The -n flag can be used to request that the resource name in the token is parsed out and displayed; exit code unchecked (exit code 2) is returned on exit. The -l flag can be used to request the number of seconds until expiry of the token is output; exit code unchecked (exit code 2) is returned on exit.

The program will only return exit codes valid or invalid if the -c flag is used, and all of the options -b bits, -d, -r resource are used. These are the minimum set of options necessary to fully check the validty of a token. If these criteria are not met, the program will return exit code unchecked (exit code 2) on exit. (See also the -y flag.)

Double Spending Protection

If the -d flag is used with the -c flag a database of spent tokens is kept.

By default tokens expire after 28 days, without expiry the database would grow indefinately. You can specify an alternate expiry period with the -e flag. The recommended (and default) expiry period for email is 28 days. After the expiry period amount of time, the token is anyway considered expired and may be purged from the database to save space. (See Purging Periodically vs on Next Access for how to purge tokens.)

For efficiency reasons a token is verified before it is checked in the database; if it is otherwise invalid no database activity will occur.

Note: The decision about how long the token should be considered valid is up to the verifier. If it is too short it is possible for some applications that the token will expire before arriving at the recipient (eg with email.) The suggested value of 28 days should be safe for normal email delivery delays. The choice is a trade-off between database size and risk of expiry prior to arrival, and depends on the application.

Note: Different tokens in the same database can have different validity periods, so for example tokens for different resources with different validity periods can be stored in the same database, or the recipient may change the validity period for future tokens without affecting the validity of old tokens.

Purging Periodically vs on Next Access

To purge old tokens periodically while checking tokens use the -p period option to purge no sooner than the given time period since the last purge. Purging can be used with the -k option to purge unexpired tokens also, and with the -r resource flag to purge only tokens for the given resource.

There are circumstances where it may be inconvenient to purge tokens on the next access, for example if there is a large double spend database which takes some time to purge, and the response time of the hashcash checker is important. To avoid this problem, purging can be done separately using just the -p now option to request just the purge operation. On unix for example you could call hashcash -p now in a cron job once per day, or on demand when disk was running low.

Speed Estimates

The -s flag requests measurement of how many collisions can be tested per second. No token is minted, or verified.

If the -b flag is used with this option, instead an estimate of how many seconds it would take to mint a token of the given size in bits is computed.

Notes

All informational output is printed on stderr. Minted tokens, and results of token verification and timing are printed on stdout. The quiet flag -q suppresses all informational output. The -v flag requests more informational output. The requested output, which is the only information that is output in quiet mode (when -q is specified) is printed on standard output. If stdout is a pipe, or when quiet mode is in effect the output is printed without description (ie just bits, just seconds, just resource).


OPTIONS

-c
Check the token given as an argument or on stdin for validity.

-m
Mint a token. If none of -c, -p, -s, -l, -n, -w are given it is anyway assumed that the user wishes to mint a token.

-b bits
When minting a token, request a collision of this many bits. When verifying a token require that it have a collision of at minimum this many bits, otherwise reject it.

-r resource
When minting tokens, the resource name (recipient's email address) to mint the token against can be given either with -r resource or as an argument to hashcash.

When checking tokens, the resource name (your own email address) can be given with the -r option. If the resource name is given it is checked against the resource name in the token, and if they do not match the token is rejected. Note if the resource name is not given, tokens for other resources would be accepted, and therefore hashcash returns exit code unchecked (exit code 2) on exit.

-e time
Expiry period for spent tokens. While checking tokens (using the -c flag), if the token was minted more than the specified amount of time ago, it is considered expired. If this option is not used, by default tokens expire after 28 days. The expiry period is given in seconds by default (an argument of 0 means forever). A single character suffix can be used to specify alternate units (m = minutes, h = hours, d = days, M = months, y = Y = years, and s = seconds).

If used with the -d option, the spent token and it's expiry period is recorded in the database. See the -p option for description of how to purge tokens from the database.

The following -e usage to give an implicit time resolution while minting is deprecated, use explicit -z option instead. (-z overrides -e if both are given. -e still works for backwards compatibility. If neither are given the default is 6 chars (time format: YYMMDD)).

While minting tokens, the -e flag can have an effect on the resolution of time created in the token. Without the -e option, the default resolution is days (time format: YYMMDD). Alternate formats based on range of expiry period are as follows:

Note the rounding down is based on UTC time, not local time. This can lead to initially suprising results when rounding down to eg days in time zones other than GMT (UTC = GMT). It may be clearer to understand if you use the -u option.

-z width
The -z option is for use during minting and deprecates the -e option given in combination with -m to specify an implicit time field width. Valid widths are 2,4,6,8,10 or 12 chars corresponding respectively to: YY, YYMM, YYMMDD, YYMMDDhh, YYMMDDhhmm, and YYMMDDhhmmss rounded down to the nearest year, month, day, hour, minute respectively.

Note the rounding down is based on UTC time, not local time. This can lead to initially suprising results when rounding down to eg days in time zones other than GMT (UTC = GMT). It may be clearer to understand if you use the -u option.

-g period
The -g option is for use when checking hashcash stamps with the -c option and specifies a grace period for clock skew, ie if a hashcash stamp arrives with a date in the future or in the past it will not be rejected as having a futuristic date (or as being expired) unless it is more futuristic (or has been expired for longer) than this period. The default is 2 days, which means as long as the sending system's clock is no more than 2 days ahead (or 2 days behind) of the receiving system's clock, the hashcash stamp will still be accepted.

The default units for grace period are seconds. A single character suffix can be used to specify alternate units (m = minutes, h = hours, d = days, M = months, y = Y = years, and s = seconds).

-d
Store tokens in a double spend database. If token has been seen before it will be rejected even if it is otherwise valid. The default database file is database.db in the current directory.

-f dbname
Use dbname instead of default filename for double spend database.

-p period
Purges the database of expired tokens if the given time period has passed since the last time it was purged. As a convenience -p now is equivalent to -p 0 both of which mean purge now, regardless of when the database was last purged.

If used in combination with -j resource only the tokens minted for the given resource are purged.

If used in combination with -k all tokens even un-expired tokens are purged. Can be used in combination with -t time to expire as if the current time were the given time.

-k
Use with option -p to request all tokens are purged rather than just expired ones.

-j resource
Use with option -p to request that just tokens matching the given resource name are to be purged, rather than the default which is to purge all expired tokens. If the resource name is the empty string, all tokens are matched (this is equivalent to omitting the -j option).

-s
Print timing information only, and don't proceed to create a token. If combined with -b flag print estimate of how long the requested collision size would take to compute, if -s given by itself, just prints speed of the collision finder.

-h
Print short usage information.

-v
Print more verbose informational output about the token minting or verification. (If -v is the only argument, prints the tool version number.)

-q
Batch mode. Prints no information other than output. This option overrides the -v option.

-X
When minting, prints the hashcash email X-header 'X-Hashcash: ' before the token. Without this option just the bare token is printed.

When checking, if no token is given as an argument, scans stdin for a line starting with the string 'X-Hashcash:', and uses the rest of the matching line as the token. Only the lines up to and ending at the first blank line are scanned. A blank line is the separator used to separate the headers from the body of a mail message or USENET article. This is meant to make it convenient to pipe a mail message or USENET article to hashcash on stdin.

-x string
Similar effect to -X for minting and checking except you get to choose your own header string. For example -x 'X-Hashcash: ' has the same effect as -X.

-i
When checking and using the -X or -x flag, ignore the blank line boundary between headers and body of the message, and check for collision in the body too if one is not found in the headers.

-t time
Pretend the current time is the time given for purposes of minting tokens, verifying tokens and purging old tokens from the database. Time is given in a format based on UTCTIME format YYMMDD[hh[mm[ss]]].

Time is expressed in local time by default. Use with -u flag to give time in UTC (GMT).

You can also give time relative to the current time by prefixing the argument with + or -. The default units for relative time are seconds. A single character suffix can be used to specify alternate units (m = minutes, h = hours, d = days, M = months, y = Y = years, and s = seconds).

Note: when time is expressed in local time, if there is daylight savings in your timezone, there are one or two ambiguous hours per year at the time of change from daylight savings time to normal time.

-u
Input and output absolute times in UTC (GMT) instead of local time.

-a period
Add (or subtract if number is negative) a random value from the current time before minting the token. This hides the time the token was created, which may be useful for anonymous users. Note adding (rather than subtracting) a random time may be risky if the token takes less than the added time to arrive as the recipient will reject tokens with time stamps in the future.

-n
Print resource name parsed from token being verified. Returns exit code unchecked on exit.

-l
Print number of seconds left before token expires. Returns exit code unchecked on exit.

Note: the calculation includes the grace period, so can be up to 2 times grace period longer than you might otherwise expect (clock fast but system has to presume it could be slow). If you want to exclude the grace period add -g0 to set grace period to 0 for the calculation.

-w
Print number of bits of collision of token. Returns exit code unchecked on exit.

-y
Returns success if the token is valid even if it is not fully checked. Use with -c where not all of -b, -d, -r are specified to get success exit code on valid but partially checked token. Similarly can use with -n, -l, -w with same effect.


EXAMPLES

Creating tokens

hashcash -s
Print timing information about how many collisions the machine can try per second.

hashcash -s -b 32
Print how long it would take the machine to compute a 32 bit collision (but don't actually compute a collision).

hashcash
Mint a token. Will prompt for resource name and desired value (number of collision bits).

hashcash foo
Compute collision on resource foo. Will prompt desired value (number of collision bits).

hashcash foo -b 10
Compute 10 bit collision on resource foo.

hashcash -a -3d
Subtract a random time of between 0 days and 3 days to to the token's creation time. This is the same fuzz factor used by mixmaster to reduce risk of timing-correlations.

Examining Tokens

hashcash -w 0:020814:foo:4333957e84db47f6
Report the value of the token (how many bits of collision) there are. The example is a 33 bit collision, which would take on average 13 hours to create on a 400 Mhz Pentium-II.

hashcash -q -b 10 foo | hashcash -w
Create a token in batch mode, pass to hashcash on stdin to verify, have it print how many bits there were. Note: half of the time you get a token 1 bit larger; similarly with decreasing probability you can get even larger tokens.

hashcash -n 0:020814:foo:21c8cf3099cbf467
Report the resource name from the token. The resource name in the example is foo.

hashcash -l -e 30y 0:020814:foo:21c8cf3099cbf467
Report how long until the token expires if it expires in 30 years from it's creation date. (Note dates too far into the future run into the 2038 end of Epoch, which is the unix time analog of the y2k bug).

Verifying Tokens

hashcash -c 0:020814:foo:21c8cf3099cbf467
Check if the token is valid. Note as we are not checking the token in a double spend database, and did not specify a resource name or required number of bits of collision and hashcash will consider the token not fully checked, and it will report it as valid but not fully unchecked, or as invalid if there is any problem with the token.

hashcash -c -b24 0:020814:foo:21c8cf3099cbf467
Check that the value of the token is greater or equal to 24 bits. This example has 24 bit value. If you increase the requested number of bits or replace the token with one with less than 24 bit collision the token will be rejected.

hashcash -c -b24 -r foo 0:020814:foo:21c8cf3099cbf467
As above check if the token has sufficient value, but in addition check that the resource name given matches the resource name in the token.

Double Spending Prevention

The examples given in Verifying Tokens can be modified to keep a double spend database so that the same token will not be accepted twice. Note a token will only be checked in and added to the database if it is otherwise valid and fully checked (a required number of bits of collision has been specified and a resource has been specified).

hashcash -cd -b 10 -r foo 0:020814:foo:21c8cf3099cbf467
Check the token and add to double spent database if it's valid (has correct resource name and sufficient value).

hashcash -cd -b 10 -r foo 0:020814:foo:21c8cf3099cbf467
Try to double spend the token. It will be rejected as double spent.

Token Expiry

To prevent the double spend database growing indefinately, the recipient can request that tokens be no older than a specified period. After expiry old tokens can dropped from the double spend database as they will no longer be needed -- expired tokens can be rejected based purely on their old date, so the space taken by expired tokens in the double spend database can be saved without risk of accepting an expired though otherwise valid token.

The first field of the token is the UTC time since 1st January 1970. The default time format is YYMMDD, time rounded down to the nearest day. The default validity period is 28 days.

You can provide an alternative validity period with the -e option.

hashcash -cd -b 10 -e 2d -r foo 0:020811:foo:21dd87d4c9f5aae1
Try verifying an old token, the above token was created 11 Aug 2002.

We gave option -e 2d so the tokens expiry date is 2 days after creation, which is now in the past.

Note: if the creation time is expressed in the token in days, the precise creation date is the begining of the specified day in UTC time (similarly for alternate units the creation time is rounded down to the begining of the unit it is expressed in). For units in days, for example, this may mean depending on your time zone that the token appears to be considered invalid in under the specified expiry period in days relative to your relative view of what day it is, as the calculation is based on current time in UTC, and the creation time of the token is expressed in UTC time.

hashcash -cd -b 10 -r foo 0:020811:foo:21dd87d4c9f5aae1
Test whether the token is otherwise valid, apart from having expired. Omitting the -e tells hashcash that the token will never expire. An expiry period of forever can also be given explitly like this: -e 0, where an expiry period of 0 means forever.

Purging old tokens

If the -c, -d options are used together, each time a token is checked, if it is valid and all of the mandatory aspects of the token are verified (collision bits check, resource name check) then the token and it's expiry period is written to the database file. The default expiry period if an expiry period is not given explicitly with the -e option is forever (ie tokens do not expire).

First mint and then add a token:

hashcash -b 10 foo -e 1m > token
Note: we specified an expiry on minting in this example, to ensure that the token creation time is given in high enough resolution in the token that the token will not be considered expired at time of creation. (Recall the default resolution is in days, a token created with a creation time rounded down to the beginging of the day is unlikely to be considered valid 1 minute later unless you mint it at midnight UTC time.)

hashcash -cd -e 1m -b 10 -r foo < token
The token expires in 1 minute. Wait 1 minute and then explicitly request that expired tokens be purged:

hashcash -p now
Then try resubmitting the same token:

hashcash -cd -e 1m -b 10 -r foo < token
and the token will be rejected anyway as it has expired, illustrating why it was not necessary to keep this token in the database.

With the default database (the sdb format) the database contents are human readable, so you can view their contents by cating them to the terminal:

cat hashcash.db
to see that the token really is added and then after puring subsequently purged due to expiry.

Purging old tokens on Demand

As a convenience you can purge at the same time as checking tokens by using the -p option with the -c option.

hashcash -b 10 foo > token =item hashcash -cd -p now -e 1 -b 10 -r foo < token
It may be inefficient to purge tokens on every use as the entire database has to be scanned for expired tokens. By giving a time period to the -p option, you can tell hashcash to purge no more frequently than that time period since the previous purge.

For example:

hashcash -cd -p 1d -e 1 -b 10 -r foo < token
tells hashcash to purge any expired tokens no more than once per day.

hashcash -p 1M -j foo
tells hashcash to purge only expired tokens matching resource foo once per month.

hashcash -p now -k
tells hashcash to purge all tokens (expired and unexpired) now.


token format

ver:date:resource:collision

where

ver = 0, this version of hashcash only supports version 0
date = YYMMDD[hh[mm[ss]]]
resource = resource string (eg IP address, email address)
collision = printable string of random alpha-numeric chars


FILES

hashcash.db
default double spend database

hashcash.dbt
default temporary double spend database used when purging tokens.


EXIT STATUS

hashcash returns success (exit code 0) after successfully minting a token, after fully checking a token and finding it valid, and after a timing test.

If when checking a token it is found to be invalid (due to being malformed, being expired, having insufficient value, having a date in the future, or being double spent), hashcash returns failure (exit code 1).

If insufficient options are given to fully check a token, or if using the -n, -l, or -w options, if the token is otherwise valid return unchecked (exit code 2). If the -y flag is given and hashcash would normally return unchecked, exit code success is returned instead.

If any exception occurs (file read failure for database checking or corrupted database contents) an exit status of 3 is returned.


AUTHOR

Written by Adam Back <adam@cypherspace.org>


SEE ALSO

sha1(1), http://www.hashcash.org/