Encrypting a short string?

[erikcw]

Hi,

I'm trying to devise a scheme to encrypt/obfuscate a short string that
basically contains the user's username and record number from the
database. I'm using this encrypted string to identify emails from a
user. (the string will be in the subject line of the email).

I'm trying to figure out which approach I should use to encrypt the
data. The string will be less than 20 characters long, and I'd like
the encrypted version to be about the same size.

I tried DES in the Crypto module, but the cipher text was to long to
be usable in this case.

Any suggestions?

Thanks!

 

[marek.rocki]

erikcw napisal(a):

Hi,

I'm trying to devise a scheme to encrypt/obfuscate a short string that
basically contains the user's username and record number from the
database. I'm using this encrypted string to identify emails from a
user. (the string will be in the subject line of the email).

I'm trying to figure out which approach I should use to encrypt the
data. The string will be less than 20 characters long, and I'd like
the encrypted version to be about the same size.

I tried DES in the Crypto module, but the cipher text was to long to
be usable in this case.

Any suggestions?

Thanks!

How about:

hashlib.sha256("(e-mail address removed)|2937267834").hexdigest()[:20]

Regards,
Marek

 

[erikcw]

erikcw napisal(a):

Hi,

I'm trying to devise a scheme to encrypt/obfuscate a short string that
basically contains the user's username and record number from the
database. I'm using this encrypted string to identify emails from a
user. (the string will be in the subject line of the email).

I'm trying to figure out which approach I should use to encrypt the
data. The string will be less than 20 characters long, and I'd like
the encrypted version to be about the same size.

I tried DES in the Crypto module, but the cipher text was to long to
be usable in this case.

Any suggestions?

Thanks!

How about:

hashlib.sha256("(e-mail address removed)|2937267834").hexdigest()[:20]

Regards,
Marek

Thanks Marek,

But that can't be reversed, right? I'd like to be able to decrypt the
data instead of having to store the hash in my database...

 

[Paul Rubin]

database. I'm using this encrypted string to identify emails from a
user. (the string will be in the subject line of the email).

1. I hope you're not trying to spam anyone.
2. What happens if the user edits the subject line?

I'm trying to figure out which approach I should use to encrypt the
data. The string will be less than 20 characters long, and I'd like
the encrypted version to be about the same size.

Under normal security requirements you cannot do this. The ciphertext
has to be longer than the plaintext since you don't want the opponent
to be able to tell whether two plaintexts are the same. Therefore you
have to attach some random padding to each plaintext. Also, you
presumably want the ciphertext to be encoded as printing characters,
while normally you'd treat the input as binary, so there is some
further expansion.

 

[marek.rocki]

erikcw napisal(a):

But that can't be reversed, right? I'd like to be able to decrypt the
data instead of having to store the hash in my database...

In such case it seems you have no choice but to use a symmetric
encryption algorithm - in other words, your original method. If the
strings are ~20 bytes long (3 DES blocks), then the base64-encoded
ciphertext will have 32 characters. In case of AES, that'll be up to
45 characters. Wouldn't such length be acceptable?

Paul Rubin napisal(a):

2. What happens if the user edits the subject line?
Under normal security requirements you cannot do this. The ciphertext
has to be longer than the plaintext since you don't want the opponent
to be able to tell whether two plaintexts are the same. Therefore you
have to attach some random padding to each plaintext. Also, you
presumably want the ciphertext to be encoded as printing characters,
while normally you'd treat the input as binary, so there is some
further expansion.

If what erikcw is looking for is a cryptographically secure protocol,
there are more things to be careful about, like authentication or
replay attacks. But indeed, I'm wondering now what his use-case is.

I'm using this encrypted string to identify emails from a
user. (the string will be in the subject line of the email).

Why not use "From" field to identify emails from a particular user?

Regards,
Marek

 

[erikcw]

erikcw napisal(a):> But that can't be reversed, right? I'd like to be able to decrypt the

In such case it seems you have no choice but to use a symmetric
encryption algorithm - in other words, your original method. If the
strings are ~20 bytes long (3 DES blocks), then the base64-encoded
ciphertext will have 32 characters. In case of AES, that'll be up to
45 characters. Wouldn't such length be acceptable?

Paul Rubin napisal(a):> 2. What happens if the user edits the subject line?

If what erikcw is looking for is a cryptographically secure protocol,
there are more things to be careful about, like authentication or
replay attacks. But indeed, I'm wondering now what his use-case is.> I'm using this encrypted string to identify emails from a

Why not use "From" field to identify emails from a particular user?

Regards,
Marek

In essence what I'm doing is trying to manage tickets for a helpdesk.
I want the ticket identifier to be short enough to fit in the subject
line along with the normal subject chosen by the user. So
cryptographic security isn't really important. I can't use the from:
field because a single user could have multiple tickets.

 

[Martin Marcher]

Hi,

In essence what I'm doing is trying to manage tickets for a helpdesk.
I want the ticket identifier to be short enough to fit in the subject
line along with the normal subject chosen by the user. So
cryptographic security isn't really important. I can't use the from:
field because a single user could have multiple tickets.

I've always wondered why such systems don't use the Message-ID or
Reference headers - I know they aren't preserved by all mailers but I
think that having this info in the subject line is

a) visually disturbing (subjective)
b) I guess that the risk of a user modifying the subject line is the
same than finding a programm that doesn't to some extent honor the
headers i mentioned...
<flame>
c) Personally whenever I find a mail that says please keep this in the
subject I delete that number on purpose...
</flame>

martin
--
http://noneisyours.marcher.name
https://twitter.com/MartinMarcher
http://www.xing.com/profile/Martin_Marcher
http://www.linkedin.com/in/martinmarcher

You are not free to read this message,
by doing so, you have violated my licence
and are required to urinate publicly. Thank you.

 

[Gabriel Genellina]

In essence what I'm doing is trying to manage tickets for a helpdesk.
I want the ticket identifier to be short enough to fit in the subject
line along with the normal subject chosen by the user. So
cryptographic security isn't really important. I can't use the from:
field because a single user could have multiple tickets.

And you don't like [bug12345] or even [12345]? To the user, it's a lot
clear its purpose, and anybody will understand what you mean if you say
"Please maintain the bug number in the subject line" or similar.

 

[Carl Banks]

In essence what I'm doing is trying to manage tickets for a helpdesk.
I want the ticket identifier to be short enough to fit in the subject
line along with the normal subject chosen by the user. So
cryptographic security isn't really important. I can't use the from:
field because a single user could have multiple tickets.

Shouldn't you have a database associating a ticket ID with an email
address (among other things)?


Carl Banks

 

[Paul Rubin]

In essence what I'm doing is trying to manage tickets for a helpdesk.
I want the ticket identifier to be short enough to fit in the subject
line along with the normal subject chosen by the user.

I think you should use a database to maintain the email addresses
since you already have to maintain the contents and history of the
help ticket anyway. If the contents of the database is private, then
assign the ticket numbers in an unpredictable sequence--I can tell you
how to do that cryptographically if you want (I've posted code for it
a few times before). That is to stop users from guessing ticket
numbers that are valid but belong to other users. If it's a public
database (e.g. a bug tracker for a free software project) or if
accessing a particular ticket needs a user credential associated with
that ticket, then you may as well use sequential numbers.

 

[Lie]

Hi,

I'm trying to devise a scheme to encrypt/obfuscate a short string that
basically contains the user's username and record number from the
database.  I'm using this encrypted string to identify emails from a
user. (the string will be in the subject line of the email).

I'm trying to figure out which approach I should use to encrypt the
data.  The string will be less than 20 characters long, and I'd like
the encrypted version to be about the same size.

I tried DES in the Crypto module, but the cipher text was to long to
be usable in this case.

Any suggestions?

Thanks!

There is a simple encryption, called ROT13 (Rotate 13). This is very
unsecure for any cryptographical purpose, but enough to make
uninformed user to think it's just a random piece of letters.

The ROT13 is done by adding 13 to each character, so
A => N,
B => O,
C => P,
D => Q, etc

the neat trick to this encryption is the algorithm is really simple
and you don't need a separate decoding algorithm as text ==
ROT13(ROT13(text)). This algorithm also guarantees that any two
different text would have two different ciphertext

 

[Bjoern Schliessmann]

There is a simple encryption, called ROT13 (Rotate 13). This is
very unsecure for any cryptographical purpose,

For enhanced security use TROT13 (triple ROT13).

but enough to make uninformed user to think it's just a random
piece of letters.

Security by obscurity doesn't work. If it needs to be protected,
protect it well. If it doesn't need to, you don't need to obscure
it at all.

Regards,


Björn

 

[Brian]

Hi Erik,

I really don't recommend the ROT13 cipher, as this is extremely easy to
crack. Most grade school kids could break this one in seconds. ;-)

If the project that you are working upon has low security needs, (in
other words, it's not a financial institution), than you might try
something quite basic such as a Vigenere, transposition, or even a
playfair. These encryption methodologies are not secure and can be
cracked by hackers. Yet, for the average Joe, it will keep them away
from the information / data stored inside.

One thing that seems to work well is to use to ciphers together. For
example, encrypt the data using the playfair cipher -- and then run a
transposition cipher. This will erase most of the "data signatures"
that are needed for most hackers to crack the code. At this point,
brute force is what most people have to resort upon -- and it's mostly
"governments" that have this ability. ;-)

Best wishes!

Dusty

 

[Steven D'Aprano]

Hi Erik,

I really don't recommend the ROT13 cipher, as this is extremely easy to
crack. Most grade school kids could break this one in seconds. ;-)

I think you missed the point. Any recommendation to use ROT13 is likely
to be a joke. A recommendation to use Triple ROT13 is *absolutely* a joke.

If the project that you are working upon has low security needs, (in
other words, it's not a financial institution),

That's rubbish. Do you think that banks are the only companies that care
about the security of their data? How would you feel if your doctor
accidentally published your medical records onto the Internet, so
everybody can see the embarrassing social diseases you've got?

(Disclaimer: I don't have any reason to think Brian actually has any
embarrassing social diseases, or any other diseases for that matter. I
was just making a rhetorical point.)

than you might try
something quite basic such as a Vigenere, transposition, or even a
playfair. These encryption methodologies are not secure and can be
cracked by hackers. Yet, for the average Joe, it will keep them away
from the information / data stored inside.

It's an awful lot of work to do those when it is so simple to call an
already existing library that is much stronger. Why do a lot of work to
get an insecure result, when you can do a little bit of work to get a
secure result?

I don't recommend any of these obsolete ciphers except as a learning
exercise, or possibly as a challenge: e.g. you *want* people to crack the
cipher, but you don't want it too easy for them.

One thing that seems to work well is to use to ciphers together. For
example, encrypt the data using the playfair cipher -- and then run a
transposition cipher. This will erase most of the "data signatures"
that are needed for most hackers to crack the code. At this point,
brute force is what most people have to resort upon -- and it's mostly
"governments" that have this ability. ;-)

That's absolute rubbish.

(1) There are well-understood techniques for breaking all these ciphers,
individually or in combination. Often, putting two of them together
doesn't make the encryption any harder to break than just using one of
them.

(2) It's not just "governments" who can break ciphers by brute force.
Anyone can do it. I could probably write a Python function to crack any
Caesar cipher in a few minutes, and it would probably run in seconds or
minutes. More complex ciphers need more work, but it's certainly
feasible: dictionary attacks are simple. Brute force only becomes
infeasible when the key is long enough. To brute force a short enough key
is within the grasp of *pencil and paper*, if you care enough.

 

[David H Wild]

I think you missed the point. Any recommendation to use ROT13 is likely
to be a joke. A recommendation to use Triple ROT13 is *absolutely* a
joke.

ROT13 does have a legitimate use, but it's not as a cypher. It is really
the equivalent of the newspaper quiz where the answers are upside down at
the bottom of the page. By doing this you stop seeing the answers too
early.

 

[Steve Holden]

ROT13 does have a legitimate use, but it's not as a cypher. It is really
the equivalent of the newspaper quiz where the answers are upside down at
the bottom of the page. By doing this you stop seeing the answers too
early.

Of course, but ROT13 ^ (2n*1) is equivalent to ROT13 for all positive
integer n. Hence the confident assertion that "A recommendation to use
Triple ROT13 is *absolutely* a joke".

regards
Steve

 

[Roy Smith]

Of course, but ROT13 ^ (2n*1) is equivalent to ROT13 for all positive
integer n.

Why restrict that to positive integers? I believe it works for all
integers. But I do think you meant 2n+1, not 2n*1.

 

[Steve Holden]

Why restrict that to positive integers? I believe it works for all
integers. But I do think you meant 2n+1, not 2n*1.

Yes, I did. "*" and "+" are much closer in my mind than they are on the
keyboard

regards
Steve