Has anyone found two files with different sizes and same MD5 sum?

[Ramon F Herrera]

I suppose it is theoretically possible to have two different files
with the same MD5 sum?

TIA,

-Ramon

[1PW]

Post by Ramon F Herrera
I suppose it is theoretically possible to have two different files
with the same MD5 sum?
TIA,
-Ramon

Hello Ramon:

It is much more than theoretically possible. An academic group has
proven that md5 based CA certificates can be forged. Albeit using a
very sophisticated hardware setup of 200 PS3 Playstations and
specialized software. Google it to read more.

Pete

[1PW]

Post by 1PW

Post by Ramon F Herrera
I suppose it is theoretically possible to have two different files
with the same MD5 sum?
TIA,
-Ramon

It is much more than theoretically possible. An academic group has
proven that md5 based CA certificates can be forged. Albeit using a
very sophisticated hardware setup of 200 PS3 Playstations and
specialized software. Google it to read more.
Pete

If you like this sort of thing this is a good read:

<http://www.win.tue.nl/hashclash/rogue-ca/>

[David E. Ross]

Post by Ramon F Herrera
I suppose it is theoretically possible to have two different files
with the same MD5 sum?
TIA,
-Ramon

Yes. However, it is more likely to occur if the situation is contrived.
That is, someone has made an effort to create a file with the same hash
as an existing file. It is unlikely (but not impossible) to occur
through mere coincidence.

Even if MD5 is "broken", it is still quite useful for checking that an
accidental modification or corruption of a file has not occurred. I
create a hash for a very large file before sending it to another
computer. I also send a file containing that hash. At the destination
computer, I again compute the hash and compare it with the hash from the
source computer. (For an extremely large file, I use SHA1 instead of
MD5. This also is a valid use of SHA1 even if it too were "broken".)

[1PW]

Post by David E. Ross

Post by Ramon F Herrera
I suppose it is theoretically possible to have two different files
with the same MD5 sum?
TIA,
-Ramon

Yes. However, it is more likely to occur if the situation is contrived.
That is, someone has made an effort to create a file with the same hash
as an existing file. It is unlikely (but not impossible) to occur
through mere coincidence.
Even if MD5 is "broken", it is still quite useful for checking that an
accidental modification or corruption of a file has not occurred. I
create a hash for a very large file before sending it to another
computer. I also send a file containing that hash. At the destination
computer, I again compute the hash and compare it with the hash from the
source computer. (For an extremely large file, I use SHA1 instead of
MD5. This also is a valid use of SHA1 even if it too were "broken".)

Hello David:

You've stolen my heart. So few vendors take the trouble to post their
MD5/SHA-1 hashes. It takes a few more moments /now/ saving much time later.

Thank you!

Pete

[dekaden|Z]

Post by David E. Ross

Post by Ramon F Herrera
I suppose it is theoretically possible to have two different files
with the same MD5 sum?
TIA,
-Ramon

Yes. However, it is more likely to occur if the situation is contrived.
That is, someone has made an effort to create a file with the same hash
as an existing file. It is unlikely (but not impossible) to occur
through mere coincidence.
Even if MD5 is "broken", it is still quite useful for checking that an
accidental modification or corruption of a file has not occurred. I
create a hash for a very large file before sending it to another
computer. I also send a file containing that hash. At the destination
computer, I again compute the hash and compare it with the hash from the
source computer. (For an extremely large file, I use SHA1 instead of
MD5. This also is a valid use of SHA1 even if it too were "broken".)

If you merely want to check for accidental (unintended) corruption (i.
e. transfer or drive errors), just use crc32 or similar. you don't need
ccryptographic hashes for that and therefore it consumes significantly
less resources.

[1PW]

Post by dekaden|Z

Post by David E. Ross

Post by Ramon F Herrera
I suppose it is theoretically possible to have two different files
with the same MD5 sum?
TIA,
-Ramon

Yes. However, it is more likely to occur if the situation is contrived.
That is, someone has made an effort to create a file with the same hash
as an existing file. It is unlikely (but not impossible) to occur
through mere coincidence.
Even if MD5 is "broken", it is still quite useful for checking that an
accidental modification or corruption of a file has not occurred. I
create a hash for a very large file before sending it to another
computer. I also send a file containing that hash. At the destination
computer, I again compute the hash and compare it with the hash from the
source computer. (For an extremely large file, I use SHA1 instead of
MD5. This also is a valid use of SHA1 even if it too were "broken".)

If you merely want to check for accidental (unintended) corruption (i.
e. transfer or drive errors), just use crc32 or similar. you don't need
cryptographic hashes for that and therefore it consumes significantly
less resources.

I know this thread is three months old. If one downloads SlavaSoft's
freeware HashCalc 2.02, they may have all the most popular hashes in a
few seconds... CRC32, MD5, SHA-1, etc.

<http://www.slavasoft.com/hashcalc/index.htm>

Works with Linux through Wine too.

Regards,

Pete

[Jorgen Grahn]

...

Post by 1PW

Post by dekaden|Z
If you merely want to check for accidental (unintended) corruption (i.
e. transfer or drive errors), just use crc32 or similar. you don't need
cryptographic hashes for that and therefore it consumes significantly
less resources.

Sure, but MD5 is almost for free too on decently recent systems.
If you check a file on disk, I/O will be the bottleneck.

CRC-32 is also just 32 bits and (I'd suppose) unsuitable for some
other non-cryptographic uses, like for finding duplicate files by
finding duplicate hashes. http://en.wikipedia.org/wiki/Birthday_problem

Post by 1PW
I know this thread is three months old. If one downloads SlavaSoft's
freeware HashCalc 2.02, they may have all the most popular hashes in a
few seconds... CRC32, MD5, SHA-1, etc.
<http://www.slavasoft.com/hashcalc/index.htm>
Works with Linux through Wine too.

Good, free implementations of those checksums are preinstalled on
every Linux machine, so running this one in Wine emulation would be a
very strange thing to do.

/Jorgen

[unruh]

["Followup-To:" header set to alt.security.pgp.]

Post by Jorgen Grahn
...

Post by 1PW

Post by dekaden|Z
If you merely want to check for accidental (unintended) corruption (i.
e. transfer or drive errors), just use crc32 or similar. you don't need
cryptographic hashes for that and therefore it consumes significantly
less resources.

Sure, but MD5 is almost for free too on decently recent systems.
If you check a file on disk, I/O will be the bottleneck.
CRC-32 is also just 32 bits and (I'd suppose) unsuitable for some
other non-cryptographic uses, like for finding duplicate files by
finding duplicate hashes. http://en.wikipedia.org/wiki/Birthday_problem

Oh, nuts. for any given file it is the full 32 bits. The 16 bits is if
you search through a whole bunch of hashes and want to check any two of
them have the same hash. In using a has for testing if it is corrupted,
it is the 32 bits that are important. That is 1/4000000000. I am willing
to check in more detail ( eg use a more complex hash) once in a billion
times.

Post by Jorgen Grahn

Post by 1PW
I know this thread is three months old. If one downloads SlavaSoft's
freeware HashCalc 2.02, they may have all the most popular hashes in a
few seconds... CRC32, MD5, SHA-1, etc.
<http://www.slavasoft.com/hashcalc/index.htm>
Works with Linux through Wine too.

Good, free implementations of those checksums are preinstalled on
every Linux machine, so running this one in Wine emulation would be a
very strange thing to do.
/Jorgen