MID Policing; RE: Parse error at the message level

I'm tending to option 2.
Because "security" may have a higher priority than "message decoder
robustness" in the majority of H.248 Control Association deployment
scenarios.

Elad, there is an entire chapter in Draft H.Sup7 on
12	H.248 Control Association: policing of incoming H.248 messages
http://www.itu.int/md/meetingdoc.asp?lang=en&parent=T05-SG16-080422-TD-W
P2-0553


The "fully secure H.248 domain" is subject of 12.3.5.
However, the general cases are discussed before.
The pre-condition of "MID policing" is: the H.248 sender must be able to
encode a correct H.248 message format with a correct MID.

-Albrecht


> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Kevin Boyle
> Sent: Dienstag, 26. Februar 2008 22:06
> To: Elad Chomsky; [email protected]
> Subject: Re: [Megaco] Parse error at the message level
> 
> Keep in mind that using Option 1 makes DoS attacks very easy 
> -- I flood your port with short malformed messages, and make 
> you generate responses to each and every one.  Not only do I 
> waste your time in processing the garbage sent, but now I 
> make you waste the processing time in generating a response 
> and an unwitting accomplice by doubling the number of 
> messages on the network.  Further, by doing this I can make 
> you send response messages that are much larger than the 
> garbage I send in, meaning that the responder becomes the one 
> that congests the network the most.
> 
> I would be extremely careful about responding to malformed 
> messages that don't even get to a TransactionID.  The more 
> compliant a message is to the protocol, the more effort 
> required to execute an attack and the less likely it will happen.
> 
> Kevin
> 
> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Elad Chomsky
> Sent: Tuesday, February 26, 2008 12:34 PM
> To: [email protected]
> Subject: [Megaco] Parse error at the message level
> 
> Hello All,
> 
> I was wondering about the correct behavior when a parse error 
> at the Message level is encountered (i.e. before any 
> transaction request or reply was parsed).
> 
> For example, consider the following message:
> 
>     !/3 [XXX 123.123.123.4]:55555
>     T = 5 {
>       ...
>     }
> 
> This message should fail parsing, as it contains an illegal mId.
> 
> 
> Option 1 is to reply with a message containing only an error 
> descriptor.
> Error code #410 seems the most appropriate. It should look 
> something like this:
> 
>     MEGACO/3 [124.124.124.222]:55555
>     Error = 410 { "Incorrect identifier: [XXX 123.123.123.4]" }
> 
> 
> Option 2 is to simply ignore the offending message and not 
> generate any reply. One (weak) corroboration to this approach 
> is that the definition of error #410 is "The transaction 
> request(s) has been disregarded due to a syntax error..."; 
> and when parsing is aborted it is not known whether the 
> message contains any transaction requests.
> 
> 
> My personal preference is option 1.
> 
> Thanks,
> Elad
> 
> _______________________________________________
> Megaco mailing list
> [email protected]
> http://www.ietf.org/mailman/listinfo/megaco
> _______________________________________________
> Megaco mailing list
> [email protected]
> http://www.ietf.org/mailman/listinfo/megaco
>