Re: [Detnet] I-D Action: draft-ietf-detnet-security-14.txt

[Yaron Sheffer <[e-mail]>]

Hi Ethan,

On the contrary. The current text is incorrect. It currently stands:


   If crypto keys are to be regenerated over the duration of the flow
   then the time required to accomplish this must be accounted for in
   the latency calculations.  Unfortunately, key generation is a
   cryptographic operation that is frequently not possible to implement
   in constant time, most notably (though not exclusively) for RSA key
   pairs.

I suggest to change to:


   If crypto keys are to be regenerated over the duration of the flow
   then the time required to accomplish this must be accounted for in
   the latency calculations.  Implementers should take care to use constant-time operations for key generation. This is normally afforded by most common key exchange protocols (e.g. those based on ECDH exchanges). Implementers should avoid the rare cases (RSA ref and specific SSH kex method here) that make constant time implementation difficult or impossible.

Thanks,
	Yaron

On 2/12/21, 01:42, "Ethan Grossman" <[e-mail]> wrote:

    Thanks Ben, I did find RFC 4432, but all I can see that it says around this
    is "This may be a transient key generated solely for
       this SSH connection, or it may be re-used for several connections." It
    doesn't say anything about being "safest to generate a key each time" so I
    figured that wasn't the right reference. Anyway I don't mean to be pedantic
    about this, I'm just trying to make the most out of the material that you've
    given me. If it isn't worth pursuing, I'm OK with leaving the text as it is
    below, if you and Yaron are OK with that. 
    Thanks again for your help...
    Ethan.

    -----Original Message-----
    From: Benjamin Kaduk <[e-mail]> 
    Sent: Thursday, February 11, 2021 3:29 PM
    To: Ethan Grossman <[e-mail]>
    Cc: 'Yaron Sheffer' <[e-mail]>; [e-mail];
    [e-mail]; [e-mail]; 'Magnus Westerlund'
    <[e-mail]>; 'Murray Kucherawy' <[e-mail]>;
    'Roman Danyliw' <[e-mail]>; 'Robert Wilton' <[e-mail]>; 'Barry
    Leiba' <[e-mail]>; 'Eric Vyncke (evyncke)'
    <[e-mail]>
    Subject: Re: [Detnet] I-D Action: draft-ietf-detnet-security-14.txt

    Hi Ethan,

    On Thu, Feb 11, 2021 at 02:39:59PM -0800, Ethan Grossman wrote:
    > Hi Ben and Yaron,
    > Thanks for this info. Does my proposed text below correctly incorporate
    your points? 
    > 
    > Regarding Ben's statement "SSH provides an "rsa2048-sha256" key exchange
    method wherein the server sends an RSA public key and the client encrypts
    some material to that key.  The guidance in that document is a little vague,
    saying that it's safest to generate a new key each time, but allowing for
    some unspecified level of reuse."  I can't seem to find the RFC for that
    reference - I got as close as
    https://tools.ietf.org/id/draft-ietf-curdle-ssh-kex-sha2-09.html
    (section 3.29) but no closer. If you could point me in the right direction I
    could update my text. 

    Sorry!  The spec for rsa2048-sha256 is RFC 4432.


    > Thanks,
    > Ethan.
    > 
    > OLD: 
    > If crypto keys are to be regenerated over the duration of the flow
    >    then the time required to accomplish this must be accounted for in
    >    the latency calculations.  Unfortunately, key generation is a
    >    cryptographic operation that is frequently not possible to implement
    >    in constant time, most notably (though not exclusively) for RSA key
    >    pairs.
    > 
    > NEW: 
    > When protocols such as IPsec/IKEv2 and MACsec are used, keys are 
    > typically only generated when a flow is instantiated, so the time

    I think I may have written sloppily to give you this impression, but it's
    not the case (at least for IPsec/IKEv2) -- IKEv2 will periodically re-key a
    connection after some amount of traffic.  It's just that IKEv2 only has
    Diffie-Hellman or similar types of key exchange, so unpredictable RSA key
    generation times are not a factor for IKEv2 re-key.

    > required for key generation does not impact the latency of packets 
    > within a flow. However, depending on the use case, it may be desirable 
    > to regenerate fresh (ephemeral) RSA keys during the flow, for example 
    > based on a time period or amount of data sent. For example, for SSH, 
    > [ref to
    > https://tools.ietf.org/html/rfc4253#section-9 ]) recommends that "keys 
    > be changed after each gigabyte of transmitted data or after each hour 
    > of connection time, whichever  comes sooner". That same reference 
    > describes the SSH process for key re-exchange, which is essentially 
    > the same mechanism as for initial key exchange.

    It also seems like this paragraph is starting to get pretty long and it's
    not clear that it's adding enough value to justify that length.
    IIUC the key point is just that rekeying during a connection is a normal and
    recommended operation, with the need to account for the delay introduced by
    key generation being pushed off to the following paragraph.
    Generating RSA keys is only noteworthy for its non-constant-time-ness, which
    is not actually mentioned here, so mentioning RSA keygen seems to be not
    doing much for us.

    > If the keys are to be regenerated over the duration of a flow then the 
    > time required to accomplish this must be accounted for in the latency 
    > calculations for that flow.  Fortunately, modern ECDH (Elliptical 
    > Curve
    > Diffie-Hellman) operations (such as x25519, e.g. see [ref to 
    > https://tools.ietf.org/html/rfc7748] ) can be performed in constant time.

    (Maybe say explicitly that since it's constant time that it can be reliably
    corrected for?)

    > END

    I think my proposal would be to stick closer to the original, with something
    like:

    In the general case cryptographic hygeine requires the generation of new
    keys during the lifetime of an encrypted flow [ref ssh], and key generation
    (or key exchange) takes additional time that must be accounted for in the
    latency calculations for that flow.  For modern ECDH key-exchange operations
    (such as x25519 [ref]), the operation can be performed in (predictable)
    constant time, though this is not universally true (such as for legacy RSA
    key exchange [ref 4432]).

    -Ben

    > -----Original Message-----
    > From: Benjamin Kaduk <[e-mail]>
    > Sent: Wednesday, February 10, 2021 8:13 PM
    > To: Ethan Grossman <[e-mail]>
    > Cc: 'Yaron Sheffer' <[e-mail]>; [e-mail]; 
    > [e-mail]; [e-mail]; 'Magnus Westerlund' 
    > <[e-mail]>; 'Murray Kucherawy' 
    > <[e-mail]>; 'Roman Danyliw' <[e-mail]>; 'Robert Wilton' 
    > <[e-mail]>; 'Barry Leiba' <[e-mail]>; 'Eric 
    > Vyncke (evyncke)' <[e-mail]>
    > Subject: Re: [Detnet] I-D Action: draft-ietf-detnet-security-14.txt
    > 
    > Interestingly enough, I can, but only as a result of doing AD review of a
    different document yesterday!
    > 
    > The point is a good one to raise, as it is definitely unusual to be doing
    ephemeral RSA key generation for key update on a live connection.
    > 
    > That said, I believe that SSH provides an "rsa2048-sha256" key exchange
    method wherein the server sends an RSA public key and the client encrypts
    some material to that key.  The guidance in that document is a little vague,
    saying that it's safest to generate a new key each time, but allowing for
    some unspecified level of reuse.  Per
    https://tools.ietf.org/html/rfc4253#section-9, key update of a live
    connection basically just uses the same mechanisms as initial key exchange,
    which thus might include generating a fresh ephemeral RSA key.
    > 
    > (I don't believe that IPsec/IKEv2 has ever supported such a mechanism, 
    > and I'm not very familiar with the details of what MACsec does, but I 
    > don't expect it to have anything like this either.)
    > 
    > My understanding matches Yaron's that one of the myriad benefits of 
    > (e.g.)
    > x25519 ECDH key exchange is that it's super-easy to implement in constant
    time.
    > 
    > -Ben
    > 
    > On Wed, Feb 03, 2021 at 11:17:21AM -0800, Ethan Grossman wrote:
    > > Thanks Yaron, glad to hear it. Regarding the RSA key pair question,
    perhaps Benjamin Kaduk could provide us some thoughts on this, since he did
    have comments around that section? Benjamin? 
    > > Ethan.
    > > 
    > > -----Original Message-----
    > > From: Yaron Sheffer <[e-mail]>
    > > Sent: Wednesday, February 3, 2021 10:58 AM
    > > To: [e-mail]; [e-mail]; [e-mail]; 
    > > [e-mail]
    > > Cc: 'Magnus Westerlund' <[e-mail]>; 'Murray 
    > > Kucherawy' <[e-mail]>; 'Roman Danyliw' <[e-mail]>; 
    > > 'Benjamin Kaduk' <[e-mail]>; 'Robert Wilton' 
    > > <[e-mail]>; 'Barry Leiba' <[e-mail]>; 'Eric
    Vyncke (evyncke)'
    > > <[e-mail]>
    > > Subject: Re: [Detnet] I-D Action: draft-ietf-detnet-security-14.txt
    > > 
    > > Thank you, this is a lot better!
    > > 
    > > A quick comment: in Sec. 7.5.1 you mention RSA key pairs. I'm not
    familiar with cases where RSA key pairs are generated on the fly as part of
    a key exchange protocol. And AFAIK (someone might want to correct me here)
    the more common, more modern ECDH operations can be performed in constant
    time.
    > > 
    > > Thanks,
    > > 	Yaron
    > > 
    > > On 2/2/21, 01:55, "Ethan Grossman" <[e-mail]> wrote:
    > > 
    > >     Hi All,
    > >     We have submitted draft 14 of the DetNet Security Considerations
    draft,
    > >     which includes resolutions for all remaining AD review comments (as
    noted in
    > >     my previous emails with the per-reviewer dispositions). 
    > >     Specifically, this version addresses comments from Yaron Sheffer,
    Magnus
    > >     Westerlund, Murray Kucherawy, Eric Vyncke, Roman Danyliw, Benjamin
    Kaduk,
    > >     Robert Wilton, and Barry Leiba.
    > > 
    > >     Thank you all for your reviews, and I hope we have addressed each of
    your
    > >     comments to your satisfaction - if you have any further comments,
    > >     suggestions or corrections please don't hesitate to let us know. 
    > > 
    > >     Sincerely,
    > >     Ethan (as Editor, DetNet Security Considerations draft)
    > > 
    > >     -----Original Message-----
    > >     From: detnet <[e-mail]> On Behalf Of
    [e-mail]
    > >     Sent: Monday, February 1, 2021 3:44 PM
    > >     To: [e-mail]
    > >     Cc: [e-mail]
    > >     Subject: [Detnet] I-D Action: draft-ietf-detnet-security-14.txt
    > > 
    > > 
    > >     A New Internet-Draft is available from the on-line Internet-Drafts
    > >     directories.
    > >     This draft is a work item of the Deterministic Networking WG of the
    IETF.
    > > 
    > >             Title           : Deterministic Networking (DetNet) Security
    > >     Considerations
    > >             Authors         : Ethan Grossman
    > >                               Tal Mizrahi
    > >                               Andrew  J. Hacker
    > >     	Filename        : draft-ietf-detnet-security-14.txt
    > >     	Pages           : 59
    > >     	Date            : 2021-02-01
    > > 
    > >     Abstract:
    > >        A DetNet (deterministic network) provides specific performance
    > >        guarantees to its data flows, such as extremely low data loss
    rates
    > >        and bounded latency (including bounded latency variation, i.e.
    > >        "jitter").  As a result, securing a DetNet requires that in
    addition
    > >        to the best practice security measures taken for any
    mission-critical
    > >        network, additional security measures may be needed to secure the
    > >        intended operation of these novel service properties.
    > > 
    > >        This document addresses DetNet-specific security considerations
    from
    > >        the perspectives of both the DetNet system-level designer and
    > >        component designer.  System considerations include a taxonomy of
    > >        relevant threats and attacks, and associations of threats versus
    use
    > >        cases and service properties.  Component-level considerations
    include
    > >        ingress filtering and packet arrival time violation detection.
    > > 
    > >        This document also addresses security considerations specific to
    the
    > >        IP and MPLS data plane technologies, thereby complementing the
    > >        Security Considerations sections of those documents.
    > > 
    > > 
    > >     The IETF datatracker status page for this draft is:
    > >     https://datatracker.ietf.org/doc/draft-ietf-detnet-security/
    > > 
    > >     There are also htmlized versions available at:
    > >     https://tools.ietf.org/html/draft-ietf-detnet-security-14
    > >     
    > > https://datatracker.ietf.org/doc/html/draft-ietf-detnet-security-14
    > > 
    > >     A diff from the previous version is available at:
    > >     https://www.ietf.org/rfcdiff?url2=draft-ietf-detnet-security-14
    > > 
    > > 
    > >     Please note that it may take a couple of minutes from the time of
    submission
    > >     until the htmlized version and diff are available at tools.ietf.org.
    > > 
    > >     Internet-Drafts are also available by anonymous FTP at:
    > >     ftp://ftp.ietf.org/internet-drafts/
    > > 
    > > 
    > >     _______________________________________________
    > >     detnet mailing list
    > >     [e-mail]
    > >     https://www.ietf.org/mailman/listinfo/detnet
    > > 
    >  
    >