On Tue, Jan 5, 2010 at 11:29 AM, Sylvain Munaut <[email protected]> wrote:
>
>
>
> First some little reminders that should shed some light :
>
> The first channel is established something like this :
>   * Paging request (BTS -> MS on PCH)
>   * Req channel (MS -> BTS on RACH)
>   * Assignement (BTS -> MS on CCCH)
>   * Paging response (MS -> BTS on the assigned channel)
>   * Cipher mode command (BTS -> MS)
>   * Cipher mode complete (MS -> BTS [ciphered!])
>   * ....
>
> The first assignement command establish the first dedicated channel.
>  * This _can_ be the TCH that will be used for the call (Very Early
> assignement).
>  * Or this can be a SDCCH
>  * This can already use hopping, but since the assignement is in clear, you
> have everything you need to 'follow' it.
>
> If a SDCCH was assigned and the phone needs a TCH (so for a voice call),
> then a new different assignement will be done and this one will take place
> on the ciphered channel so you can't see it ...
>
>
>>
>> > the computational complexity does not increase because of that.
>> > what happens is that you break the cipher with the few frames
>> > that are sent on the original SDCCH after encryption is switched
>> > on and before a new hopping sequence is negotiated. then you know
>> > which frames to decrypt with the now known key from the wideband
>> > capture.
>>
>> Hum, this contradicts with what I've heard from other people, but
>> I didn't read this part of specs by myself and can't judge.
>
> See above.
>
> yes you could break encryption with just the first few frames on the
> original SDCCH (which is known) but that _will_ need more computational
> complexity because you have less plain text -> less chance of finding the
> key.
>
>
>> > 5) You could try to 'brute force' the hopping parameters by just looking
>> > at
>> > the enery in each timeslot at given time. For a given cell, only the
>> > 'timeslot' & 'index' changes. But that would need FPGA code changes to
>> > just
>> > get the energy back and not the samples themselves.
>>
>> hum. I thought that hopping sequence is generated using encryption key
>> as a parameter. Am I wrong and sequence is the same, and only index in
>> the sequence is changed?
>
> The hopping has nothing to do with ciphering. It just happens that the
> second 'assignement' command that carries what you need to generate the
> sequence for the TCH is transfered ciphered.
>
> To know where to listen you need some parameters that are in the
> 'assignment' command :
>  - The time slot (doesn't change during hopping)
>  - The list of channels used by that cell and the subset used for hopping
>  - Hopping Sequence Number (HSN)
>  - The 'index' (integer between 0 and N-1)
>
> Some of theses are constant for a given cell (can't really be any other way
> if you want to avoid collisions, and observation on real network confirm
> this) :
>  - HSN
>  - subset of the list of channels used for hopping (usually all but C0)
> so if you place a call in that cell yourself, you can get theses easily
> before hand. The only remaining unknowns are the timeslot and the index.
>
>>
>> Also this will work only if at every point of time there is a single
>> active phone
>
> No, you can eliminate channels that had energy before hand. So you need a
> single 'new' phone at a time. But granted it's a little far fetched and
> would not be the best method for the usrp because the LO would take too much
> time to settle anyway. But if you can get the whole spectrum, you can
> bruteforce it to try and get more plaintext (the SI5/SI6 messages). This
> decouples the problem of getting plain text from getting the hopping
> sequence.
>
>
>     Sylvain
>
>
>
> _______________________________________________
> A51 mailing list
> [email protected]
> http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
>
>

Hi,
Thanks for good summary Sylvain, for everyone interested the ETSI spec
name describing this is 05.02.

http://www.3gpp.org/ftp/Specs/html-info/0502.htm

Regards Kugg
_______________________________________________
A51 mailing list
[email protected]
http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51

Reply via email to