On 28 Dec 2003 05:51:56 -0800, gohyongkwang@hotmail.com (Goh, Yong
Kwang) wrote, in part:

I would like to find out if I apply base 64 encoding to a binary
file/stream before applying symmetric encryption, will the resulting
ciphertext be easier to break and cryptanalyzed?

Yes and no.

Yes, in general, because the limitation to 64 symbols creates
redundancy, which can be used in cryptanalysis.

No, in particular, because modern block ciphers have key sizes immune
to brute-force search, and are secure against known-plaintext attack.

But it's a bad idea anyways, because it leads to having to use the
encryption algorithm more times, thus it leads to inefficiency.

I would like to find out if I apply base 64 encoding to a binary
file/stream before applying symmetric encryption, will the resulting
ciphertext be easier to break and cryptanalyzed?

Hi.

I would like to find out if I apply base 64 encoding to a binary
file/stream before applying symmetric encryption, will the resulting
ciphertext be easier to break and cryptanalyzed?

I've only some basic understanding of encryption and mathematics and
is not quite sure how the base 64 encoding would affect the security
of the ciphertext and encryption.

But to me it may reduce the security because base 64 encoding reduces
the number of symbols (characters) used to represent the plaintext. In
original binary mode, there would be 256 combinations for each byte.
Whereas in base 64 encoding, there would be only 65 combinations in
use for each byte, thus my rationale is that it may make it easier for
the attacker to do some statistical cryptanalysis.

Any comment will be appreciated.
Thank you.

---
Goh, Yong-Kwang
Singapore
gohyongkwang@hotmail.com

And, in fact, it's a waste of time, because you'll only have to base-64
encode it again before emailing it to the legitimate recipient. :-)

gohyongkwang@hotmail.com (Goh, Yong Kwang) writes:

]Hi.

]I would like to find out if I apply base 64 encoding to a binary
]file/stream before applying symmetric encryption, will the resulting
]ciphertext be easier to break and cryptanalyzed?

Yes. BAse64 encoding replaces the 3 8 bit bytes of binary by 4 6 bit
bytes, made up of printable characters only. Thus on decoding the
attacker need only check to make sure that each byte is one of the
base64 characters. If not, it is not the right key.
However, this is not much of an advantage. He would still have to use
exhaustive search to find the right key. It just makes it easier to
determine if he has the right key.
(Ie, the "easier" is not much)

]I've only some basic understanding of encryption and mathematics and
]is not quite sure how the base 64 encoding would affect the security
]of the ciphertext and encryption.

]But to me it may reduce the security because base 64 encoding reduces
]the number of symbols (characters) used to represent the plaintext. In
]original binary mode, there would be 256 combinations for each byte.
]Whereas in base 64 encoding, there would be only 65 combinations in
]use for each byte, thus my rationale is that it may make it easier for
]the attacker to do some statistical cryptanalysis.

[snip]
But to me it may reduce the security because base 64 encoding reduces
the number of symbols (characters) used to represent the plaintext. In
original binary mode, there would be 256 combinations for each byte.
Whereas in base 64 encoding, there would be only 65 combinations in
use for each byte, thus my rationale is that it may make it easier for
the attacker to do some statistical cryptanalysis.

"Goh, Yong Kwang" wrote:

[snip]
But to me it may reduce the security because base 64 encoding reduces
the number of symbols (characters) used to represent the plaintext. In
original binary mode, there would be 256 combinations for each byte.
Whereas in base 64 encoding, there would be only 65 combinations in
use for each byte, thus my rationale is that it may make it easier for
the attacker to do some statistical cryptanalysis.

But this (reversible) conversion doesn't involve any
encryption key. So the two forms are simply equivalent
'representations' of the same thing and thus shouldn't
inherently affect the statistical properties in them.

"Mok-Kong Shen" <mok-kong.shen@t-online.de> wrote in message
news:3FEF6DE4.8C7BA625@t-online.de...


"Goh, Yong Kwang" wrote:

[snip]
But to me it may reduce the security because base 64 encoding
reduces the number of symbols (characters) used to represent the
plaintext. In original binary mode, there would be 256 combinations
for each byte. Whereas in base 64 encoding, there would be only 65
combinations in use for each byte, thus my rationale is that it may
make it easier for the attacker to do some statistical
cryptanalysis.

But this (reversible) conversion doesn't involve any
encryption key. So the two forms are simply equivalent
'representations' of the same thing and thus shouldn't
inherently affect the statistical properties in them.

This may require a little more thought. First, a (reversible) keyed
AES encipherment of a block of 16 plaintext bytes is a "simply
equivalent representation of the same thing" and yet the statistical
properties of the ciphertext will differ wildly from those of the
plaintext.

Second, the analysis of the statistical properties of Base-64 encoding
must show *Zero* probability that one will ever see bit combinations
outside of the basic set of characters (A-Z, a-z, and a few others)
regardless of the distribution of bit patterns in the source
material--another example of a change in statistical properties by an
unkeyed reversible conversion.

Third, Base-64 encoding can be considered to be a polygraphic,
polyalphabetic substitution cipher with a fixed, known key and a
relatively short period.

Base-64 encoding adds no entropy to the plaintext, but it does
increase the amount of plaintext available and provides a signature
for cryptanalysis in the form of a characteristic probability
distribution. It is a "data armoring" technique historically used to
ensure that natively 8-bit character data can be routed through legacy
communication systems that were limited to processing 7-bit
characters.

"Mok-Kong Shen" <mok-kong.shen@t-online.de> wrote:


"Goh, Yong Kwang" wrote:

[snip]
But to me it may reduce the security because base 64 encoding reduces
the number of symbols (characters) used to represent the plaintext. In
original binary mode, there would be 256 combinations for each byte.
Whereas in base 64 encoding, there would be only 65 combinations in
use for each byte, thus my rationale is that it may make it easier for
the attacker to do some statistical cryptanalysis.

But this (reversible) conversion doesn't involve any
encryption key. So the two forms are simply equivalent
'representations' of the same thing and thus shouldn't
inherently affect the statistical properties in them.

This may require a little more thought. First, a (reversible) keyed AES
encipherment of a block of 16 plaintext bytes is a "simply equivalent
representation of the same thing" and yet the statistical properties of the
ciphertext will differ wildly from those of the plaintext.
[snip]

encoding. Each bit of the source is expanded that the value of each 1
bit is the 8 bytes 0x8000000000000000 and each 0 bit is the 8 bytes
0x00000000000000. Now the encoded text clearly has exactly the same
information as the original (it is only 64 times as long). But
cryptanalysis of this is trivial. There will only be two outputs for
each 64 bit block of the output. One of these is 1 and the other is 0.

"John E. Hadstate" wrote:

"Mok-Kong Shen" <mok-kong.shen@t-online.de> wrote:


"Goh, Yong Kwang" wrote:

[snip]
But to me it may reduce the security because base 64 encoding
reduces
the number of symbols (characters) used to represent the plaintext.
In
original binary mode, there would be 256 combinations for each byte.
Whereas in base 64 encoding, there would be only 65 combinations in
use for each byte, thus my rationale is that it may make it easier
for
the attacker to do some statistical cryptanalysis.

But this (reversible) conversion doesn't involve any
encryption key. So the two forms are simply equivalent
'representations' of the same thing and thus shouldn't
inherently affect the statistical properties in them.

This may require a little more thought. First, a (reversible) keyed AES
encipherment of a block of 16 plaintext bytes is a "simply equivalent
representation of the same thing" and yet the statistical properties of
the
ciphertext will differ wildly from those of the plaintext.
[snip]

But here (as I called attention to) a key is involved
and that key is kept unknown to the opponent. This is
a secret transformation not a 'public' transformation.
The opponent, if he is capable, will try all that is
at his disposal. A transform that is 'known' to him,
doesn't cause any trouble to him (we neglect some
resources that would be needed e.g. to back-transform).

Consider also this example: A 128-bit block can e.g. be
represented as 4 integers in hex format or in decimal
digits format. Does the one form contain more
'information' than the other? (Note that the 'base' of
the two are different.)

Of course, the opponent, in doing
his statistical analysis, may try certain different
equivalent formats and some would be more appropriate
or convenient than the others.