IETF94 SAAG Meeting Agenda (120 mins), Yokohama, Japan 13:00-15:00 Thursday Afternoon Session I Chairs: Stephen Farrell, Kathleen Moriarty (in absentia) Thanks to Wes George and Robin Winton for notes. 1. WG/BOF reports as needed (10 minutes) 2. Invited/offered talks --------- 2.1 Keychain-based YANG models in the NETCONF WG, Kent Watsen https://tools.ietf.org/html/draft-ietf-netconf-server-model-08 EKR: What are the semantics in these containers? Some datastores specify purpose, this seems unstructured Modeling on any systems? PKCS12 is an existing standard – not a good one, but existing SPT: willing to help Rich Salz: pkcs12 data model is well understood, might help Jabber room: PHB: “I wouldn’t advise it. PKCS#12 is complex…" Kathleen: RFC7292 can be updated within the IETF --------- 2.6 YANG – Key Table and Key Chain, Russ Housley (15 minutes) https://datatracker.ietf.org/doc/draft-acee-rtg-yang-key-chain/ http://datatracker.ietf.org/doc/draft-chen-rtgwg-key-table-yang/ SPT: arguing about data models. Pick them, let's move on. Goal: what are the things you want to manage? --------- 2.2 On the standardization of cryptographic application techniques for IoT devices in ITU-T and ISO/IEC JTC 1, Hirotaka Yoshida (Hitachi) (20 Minutes) Hannes T: various efforts to define new crypto, esp symmetric. Why focus on those? Perf problems aren't a big issue on symmetric, even when you increase efficiency by factor of 10, why not optimize on asymmetric where there's a problem? IoT needs lots of other functions, so crypto may not be limiting factor, protocol stack is important. Look at sites that sell chips, counterintuitive economies of scale. DKG: Speed opt that you're shooting for not the right place to optimize. Approach to select which data are sensitive and need to be encrypted is something we've failed at repeatedly. We're not got at being able to tell which are sensitive and need to be encrypted. If you're doing this, not even protecting integrity of the other pieces of date. Not about algorithm, but how you determine which is which sensitive/non-sensitive. Start by encrypting everything, profile the system, and only peel back stuff if you really need to for performance limit reasons Jari: value to optimize IoT. Research on resource usage. Power consumption – radio communications are far larger drain on power than anything else, including crypto. Work to be done for sec. standards for IoT, but more in how to reduce number of proto exchanges or bits in a packet, rather than reducing CPU. EKR: what fraction of cycles are symmetric vs asymmetric. That should be first thing to know if you want to shave cycles. --------- 2.3 A Holistic Threat Analysis of IPv6 Transition Technologies, Marius Georgescu (NAIST/WIDE) (20 Minutes) Where to do this work? V6ops? --------- 2.4 MARNEW Workshop Summary and Next Steps, Natasha Rooney (15 Minutes) Jana – want to try turning off net mgmt, but also want to try enabling AQM, which isn't there currently Michael Richardson: border skirmish, transport-friendly ESP, always asked mobile folks, great, what's in it for me that makes this worthwhile to change the spec? Where will I learn this? Ted Hardie: it's in the minutes, but radio networks decide on the resources to assign by examining the flow. Encryption makes that hard – know about the flow but not what's in it EKR: depends on what data you're exposing. Not like providers have never interfered with traffic based on what it was. What info can I tell them that won't be used against me. Ted: Loss tolerant vs non loss tolerant Natasha – looking for non-metadata-style solutions. Transport based stuff might help address this Jana – Complex space. Don't want to divide this metadata vs not. Want to talk about overall system. Mobile ops took easier approach in the past because they had access, now legacy stuff. Not even sure they know what's going on. They think they're doing the right things. Seem open to engaging. Don’t always have visibility into what the middleboxes do for end user Natasha – suggest data that you'd like to have from the mobile network Joe Hildebrand – workshop began a relationship between us vs them – working together now. Build a network that has privacy, performance, etc. Learned a lot about how the transport actually works. Really bad interactions with the transports we use in standard ways today. Considering those inputs gives us a shot to do interesting new work. Wes – re no crypto/no interception – does that include no trusted proxy, forced certs because provider owns device? Natasha – yes, a few papers on this, but not well received ---------- 2.5 No MTI without Public Review draft, Rich Salz (5 minutes) ---------- 3. Open Mic, (30 Minutes) Jana – re: marnew – exactly the conversation we wanted. Operators now forced to come to the table and have a conversation because they are losing access to info they had before PHB: look at matt blaze work on proxy reencryption SPT: TLS – trying to change when academics look at specs, get TLS1.3 spec done, then pause 6 months so that they can review. TRON (TLS1.3) ready or not Kevin Smith, Vodaphone: We have no evidence that encryption has problem. Running networks 50% encrypted. Have time to get some of these other congestion control best practices in. Wendy Seltzer: W3C – other work on Web Auth, HW-based security. Web Crypto API nearing proposed recommendation stage, encouraging specs to include security and privacy considerations Christian Huitema: things in the radio network designed to act based on inferences about what applications need, which aren't true. Example – L2 retrans. Huge jitter. Mismatch between what one team believes is needed and what is actually needed. Good conversation. Wes George - some vehicle communications/emergency drafts need a security focus. Recommended to the authors to investigate, but wanted to mention here too. draft-ietf-ecrit-ecall draft-ietf-ecrit-car-crash draft-petrescu-its-problem draft-petrescu-its-cacc-sdo