Likes Likes:  0
Resultaten 1 tot 2 van de 2
Geen
  1. #1
    Olivier
    Cisco ACL bug when using VPN crypto engine accelerator, PPPoE
    Gast
    n/a Berichten
    Berichten zijn liked



    Thread Starter

    Cisco ACL bug when using VPN crypto engine accelerator, PPPoE



    Platform Cisco 1760 dual Ethernet

    IOS 12.2.xT IP/ADSL/FW/IDS PLUS IPSEC 3DES

    Environment: Site to site VPN for small offices.



    ACL are not properly parsed as soon as you enable:

    crypto engine accelerator
    PPPoE dialer
    Ip route-cache


    Without the feature mentioned above, you can apply an ACL on the outside
    interface allowing only inbound ISAKMP and IPSEC traffic.

    I.E.

    ip access-list extended Block-Inbound-unwanted-Trafic

    permit udp 100.100.100.0 0.0.0.255 host 102.168.1.2 eq isakmp

    permit esp 100.100. 100.0 0.0.0.255 host 102.168.1.2

    deny ip any any log



    If you activate the crypto engine, the ACL is parsed as well on decrypted
    traffic which forces you to allow as well all traffic for the decrypted
    traffic.
    I.E. If you are using 10.x addressees internally and the subnet
    10.200.0.0/24 for your Soho LAN. Can be worst if you have a huge network
    inside where you would prefer to add permit ip any 10.200.0.0 0.0.0.255.


    ip access-list extended Block-Inbound-unwanted-Trafic
    permit udp 100.100.100.0 0.0.0.255 host 102.168.1.2 eq isakmp
    permit esp 100.100. 100.0 0.0.0.255 host 102.168.1.2
    permit ip 10.0.0.0 0.255.255.255 10.200.0.0 0.0.0.255 <-----------@%#$%@
    deny ip any any log


    This looks pretty bad for a VPN box running a Firewall feature set IOS
    seen as the best candidate for VPN for small offices.

    The worst is the reply from Cisco:
    -------------------------------------------------------------------
    We will be addressing this in the next few months however
    the release time frame could be as late as the end
    of the year.

    We do have plans to address it but do
    not expect it in a released image until the
    last calendar quarter of the year. If its possible we
    can get it done and released sooner than what I've
    mentioned, we will do it, no guarantees however.
    -------------------------------------------------------------------

    We would have hope that they put more resources and concern in solving
    security issue.


  2. #2
    Ilker Temir
    Cisco ACL bug when using VPN crypto engine accelerator, PPPoE
    Gast
    n/a Berichten
    Berichten zijn liked



    Thread Starter

    Re: Cisco ACL bug when using VPN crypto engine accelerator, PPPoE

    -----BEGIN PGP SIGNED MESSAGE-----
    Hash: SHA1

    This is in response to the e-mail sent by Olivier. The original e-mail is
    available at
    http://www.securityfocus.com/archive...2/2003-05-18/0

    Hi Olivier,

    We can confirm that the inbound ACL is processed twice for IPSec traffic
    which is the root cause of the issue you have described.

    This issue is being addressed by the Cisco Bug ID CSCdz54626. Cisco's
    Product Security Incident Response Team (PSIRT) was not previously aware
    of it. We are now addressing it with due priority.

    While this issue does cause administrative overhead to configuration,
    there is no significant security impact.

    Permitting internal networks in the inbound ACL may be exploited to inject
    spoofed packets into the network. However this has no practical impact
    while using static crypto maps. With static crypto maps, the unencrypted
    traffic will be dropped even if it passes the inbound ACL.

    In the case of dynamic crypto maps, in order to bypass the inbound ACL to
    inject spoofed packets, an attacker would also require control of the
    neighboring routers that are connected to the interface where the inbound
    ACL is applied, or the medium in between the neighbors.

    Protecting our customers' networks is very important for us and we are
    always open for vulnerability reports regarding any Cisco product.
    Such reports should be directly sent to psirt@cisco.com or
    security-alert@cisco.com.

    Thank you again for your report,

    Regards,

    - --
    Ilker Temir
    Incident Manager, PSIRT
    Cisco Systems, Inc.
    +32 2 704-6031
    http://www.cisco.com/go/psirt

    On Wed, 14 May 2003, Olivier wrote:

    >
    >
    > Platform Cisco 1760 dual Ethernet
    >
    > IOS 12.2.xT IP/ADSL/FW/IDS PLUS IPSEC 3DES
    >
    > Environment: Site to site VPN for small offices.
    >
    >
    >
    > ACL are not properly parsed as soon as you enable:
    >
    > crypto engine accelerator
    > PPPoE dialer
    > Ip route-cache
    >
    >
    > Without the feature mentioned above, you can apply an ACL on the outside
    > interface allowing only inbound ISAKMP and IPSEC traffic.
    >
    > I.E.
    >
    > ip access-list extended Block-Inbound-unwanted-Trafic
    >
    > permit udp 100.100.100.0 0.0.0.255 host 102.168.1.2 eq isakmp
    >
    > permit esp 100.100. 100.0 0.0.0.255 host 102.168.1.2
    >
    > deny ip any any log
    >
    >
    >
    > If you activate the crypto engine, the ACL is parsed as well on decrypted
    > traffic which forces you to allow as well all traffic for the decrypted
    > traffic.
    > I.E. If you are using 10.x addressees internally and the subnet
    > 10.200.0.0/24 for your Soho LAN. Can be worst if you have a huge network
    > inside where you would prefer to add permit ip any 10.200.0.0 0.0.0.255.
    >
    >
    > ip access-list extended Block-Inbound-unwanted-Trafic
    > permit udp 100.100.100.0 0.0.0.255 host 102.168.1.2 eq isakmp
    > permit esp 100.100. 100.0 0.0.0.255 host 102.168.1.2
    > permit ip 10.0.0.0 0.255.255.255 10.200.0.0 0.0.0.255 <-----------@%#$%@
    > deny ip any any log
    >
    >
    > This looks pretty bad for a VPN box running a Firewall feature set IOS
    > seen as the best candidate for VPN for small offices.
    >
    > The worst is the reply from Cisco:
    > -------------------------------------------------------------------
    > We will be addressing this in the next few months however
    > the release time frame could be as late as the end
    > of the year.
    >
    > We do have plans to address it but do
    > not expect it in a released image until the
    > last calendar quarter of the year. If its possible we
    > can get it done and released sooner than what I've
    > mentioned, we will do it, no guarantees however.
    > -------------------------------------------------------------------
    >
    > We would have hope that they put more resources and concern in solving
    > security issue.
    >

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.2.1 (SunOS)

    iD8DBQE+w8bR8/wE0ppYtwURAunzAJ4oUlepUBjdJzQ1jZfBQGNI3UNNkwCcCcRh
    CJBxZPAMkMO9/PwFxVcibTs=
    =4dJU
    -----END PGP SIGNATURE-----

Webhostingtalk.nl

Contact

  • Rokin 113-115
  • 1012 KP, Amsterdam
  • Nederland
  • Contact
© Copyright 2001-2026 Webhostingtalk.nl.
Web Statistics