2018-03-12 17:07:53

by Tudor Ambarus

[permalink] [raw]
Subject: in-kernel user of ecdsa

Hi,

Would you consider using ECDSA in the kernel module signing facility?
When compared with RSA, ECDSA has shorter keys, the key generation
process is faster, the sign operation is faster, but the verify
operation is slower than with RSA.

Smaller key sizes imply reduced memory footprint and bandwidth that are
especially attractive for memory constrained devices. I'm working with
such a device, capable of generating ecc keys, secure key storage and
ecdsa/ecdh crypto acceleration. I'm trying to find an in-kernel user of
ecdsa.


ECDSA and RSA comparison
------------------------
-> ECDSA requires a much smaller key length in order to provide the same
security strength as RSA [1]:

Security Strength RSA (bits) ECDSA (bits)
112 2048 224 - 255
128 3072 256 - 383
192 7680 384 - 511
256 15360 512+

7680 and 15360 keys are not included in the NIST standards for
interoperability and efficiency reasons, the keys are just too big.

-> key generation: ECC key generation is faster than IFC (Integer -
Factorization Cryptography). RSA private key is based on large prime
numbers, while for ECDSA any positive integer less than n is a valid
private key.

-> ECDSA sign operations are faster than RSA, but verify operations are
slower. Here's an openssl speed test that I've run on my computer:

sign verify sign/s verify/s
rsa 2048 bits 0.000604s 0.000018s 1656.3 56813.7
rsa 4096 bits 0.004027s 0.000062s 248.3 16052.5

sign verify sign/s verify/s
256 bit ecdsa (nistp256) 0.0000s 0.0001s 28986.4 13516.3
384 bit ecdsa (nistp384) 0.0002s 0.0008s 5541.0 1322.2
521 bit ecdsa (nistp521) 0.0003s 0.0006s 3104.2 1756.2

Best,
ta

[1] NIST SP 800-57 Pt. 1 Rev. 4, Recommendation for key management

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot


2018-03-12 18:09:18

by James Bottomley

[permalink] [raw]
Subject: Re: in-kernel user of ecdsa

On Mon, 2018-03-12 at 19:07 +0200, Tudor Ambarus wrote:
> Hi,
>
> Would you consider using ECDSA in the kernel module signing facility?
> When compared with RSA, ECDSA has shorter keys, the key generation
> process is faster, the sign operation is faster, but the verify
> operation is slower than with RSA.

You missed the keyrings list, which is where the module signing utility
is discussed.

First question is, have you actually tried?  It looks like sign-file
doesn't do anything RSA specific so if you give it an EC X.509
certificate it will produce an ECDSA signature.

I think our kernel internal x509 parsers don't have the EC OIDs, so
signature verification will fail; but, especially since we have the
rest of the EC machinery in the crypto subsystem, that looks to be
simply fixable.

James


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
tpmdd-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/tpmdd-devel

2018-03-12 21:55:35

by James Bottomley

[permalink] [raw]
Subject: Re: in-kernel user of ecdsa

On Mon, 2018-03-12 at 20:56 +0100, Stephan Mueller wrote:
> Am Montag, 12. März 2018, 19:09:18 CET schrieb James Bottomley:
>
> Hi James,
>
> >
> > On Mon, 2018-03-12 at 19:07 +0200, Tudor Ambarus wrote:
> > >
> > > Hi,
> > >
> > > Would you consider using ECDSA in the kernel module signing
> > > facility? When compared with RSA, ECDSA has shorter keys, the key
> > > generation process is faster, the sign operation is faster, but
> > > the verify operation is slower than with RSA.
> >
> > You missed the keyrings list, which is where the module signing
> > utility is discussed.
> >
> > First question is, have you actually tried?  It looks like sign-
> > file doesn't do anything RSA specific so if you give it an EC X.509
> > certificate it will produce an ECDSA signature.
> >
> > I think our kernel internal x509 parsers don't have the EC OIDs, so
> > signature verification will fail; but, especially since we have the
> > rest of the EC machinery in the crypto subsystem, that looks to be
> > simply fixable.
>
> ECDSA is not implemented currently in the kernel crypto API.

an ECDSA signature is produced as a ECDH operation using the DSA
algorithm instead of KDFe, so it's trivial with what we have; signature
verification involves a separate point addition but we have all the
primitives for this in crypto/ecc.c so adding it isn't really
difficult, is it?

James


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
tpmdd-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/tpmdd-devel

2018-03-26 14:59:09

by Tudor Ambarus

[permalink] [raw]
Subject: Re: in-kernel user of ecdsa



On 03/12/2018 07:07 PM, Tudor Ambarus wrote:

> Would you consider using ECDSA in the kernel module signing facility?

Any feedback is good. I can invest some time to make this happen, if
needed.

> When compared with RSA, ECDSA has shorter keys, the key generation
> process is faster, the sign operation is faster, but the verify
> operation is slower than with RSA.
>
> Smaller key sizes imply reduced memory footprint and bandwidth that are
> especially attractive for memory constrained devices. I'm working with
> such a device, capable of generating ecc keys, secure key storage and
> ecdsa/ecdh crypto acceleration. I'm trying to find an in-kernel user of
> ecdsa.
>
>
> ECDSA and RSA comparison
> ------------------------
> -> ECDSA requires a much smaller key length in order to provide the same
> security strength as RSA [1]:
>
> Security Strength RSA (bits) ECDSA (bits)
> 112 2048 224 - 255
> 128 3072 256 - 383
> 192 7680 384 - 511
> 256 15360 512+
>
> 7680 and 15360 keys are not included in the NIST standards for
> interoperability and efficiency reasons, the keys are just too big.
>
> -> key generation: ECC key generation is faster than IFC (Integer -
> Factorization Cryptography). RSA private key is based on large prime
> numbers, while for ECDSA any positive integer less than n is a valid
> private key.
>
> -> ECDSA sign operations are faster than RSA, but verify operations are
> slower. Here's an openssl speed test that I've run on my computer:
>
> sign verify sign/s verify/s
> rsa 2048 bits 0.000604s 0.000018s 1656.3 56813.7
> rsa 4096 bits 0.004027s 0.000062s 248.3 16052.5
>
> sign verify sign/s verify/s
> 256 bit ecdsa (nistp256) 0.0000s 0.0001s 28986.4 13516.3
> 384 bit ecdsa (nistp384) 0.0002s 0.0008s 5541.0 1322.2
> 521 bit ecdsa (nistp521) 0.0003s 0.0006s 3104.2 1756.2
>
> Best,
> ta
>
> [1] NIST SP 800-57 Pt. 1 Rev. 4, Recommendation for key management
> --
> To unsubscribe from this list: send the line "unsubscribe keyrings" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot