On Fri, Dec 14, 2018 at 02:11:21AM +0000, Mach Chen wrote: > Hi Linda, > > Thanks for the review! > > Some responses inline... > > > -----Original Message----- > > From: ietf [mailto:ietf-bounces@xxxxxxxx] On Behalf Of Linda Dunbar > > Sent: Wednesday, December 12, 2018 4:24 AM > > To: secdir@xxxxxxxx > > Cc: mpls@xxxxxxxx; draft-ietf-mpls-lsp-ping-lag-multipath.all@xxxxxxxx; > > ietf@xxxxxxxx > > Subject: Secdir last call review of draft-ietf-mpls-lsp-ping-lag-multipath-05 > > > > Reviewer: Linda Dunbar > > Review result: Ready > > > > I have reviewed this document as part of the security directorate's ongoing > > effort to review all IETF documents being processed by the IESG. These > > comments were written primarily for the benefit of the security area > > directors. Document editors and WG chairs should treat these comments > > just like any other last call comments. > > > > The summary of the review is Ready with comment > > > > The described mechanism for LSP Multipath Ping is very clear. The Security > > Consideration re-uses the description of RFC8029, which is very > > comprehensive. > > It would be better if the draft describes how to prevent intermediate LSRs in > > between the Initiating LSR and Responding LSR from mis-using the detailed > > link information (e.g. forwarding to somewhere else). > > The Echo Request and Reply messages are directly exchanged between the Initiating LSR and the Responding LSR, those intermediate LSRs just forward the messages as normal packets, they will not see the detailed link information unless if they inspect and do DPI on every packet forwarded by them. > > The detailed link information is supplied to the Initiating LSR for using, the intermediate LSRs will not try to use it even if they received the information, because there is no corresponding Echo Request to the received Echo Reply. The intermediate LSRs still will have access to the plaintext information, even if in normal operation they do not need to act upon that information. Generally in this sort of situation we will either explicitly state that the intermediate nodes must be trusted to not abuse the information in question, or provide some mechanism for end-to-end confidentiality protection. Also (noting that I only skimmed the document so this may not make sense), the security considerations seem to suggest using an IP ACL for determining which messages are trusted; IP ACLs are generally not recommended in favor of cryptographic mechanisms at this point. -Benjamin
