180 likes | 304 Views
RMT Meeting. IETF 71 st – Philadelphia, March 2008 Vincent Roca. INRIA Rhône-Alpes - V. Roca -. Outline. RMT Security TESLA for ALC/NORM status Simple Authentication Schemes RMT securityI-D: what’s next? FCAST status why should it be considered? next steps.
E N D
RMT Meeting IETF 71st – Philadelphia, March 2008 Vincent Roca INRIA Rhône-Alpes - V. Roca -
Outline • RMT Security • TESLA for ALC/NORM status • Simple Authentication Schemes • RMT securityI-D: what’s next? • FCAST • status • why should it be considered? • next steps INRIA Rhône-Alpes - V. Roca -
RMT Security Discussion INRIA Rhône-Alpes - V. Roca -
TESLA for ALC/NORM status • basically • IMHO the specification is now stable • ready for WGLC? (⇒ wednesday MSEC meeting) • we implemented most of it to cross-check the specification... • seems to work but we did not test it thoroughly, though • a serious test campaign is planned INRIA Rhône-Alpes - V. Roca -
TESLA for ALC/NORM status... (cont’) • what we did in new version, in short • new bootstrap information message format that contains a “key chain commitment” rather than disclosing a key ⇒ in line with some authentication tags • added a “crypto function type” field/IANA registry to be used with digital signatures • finished description of the steps needed to authenticate incoming packets • clarified the use of multiple key chains • etc. • See email (03/09/08) for technical details INRIA Rhône-Alpes - V. Roca -
Simple Auth Schemes for ALC/NORM • no progress... • at IETF’70 the decision was to wait feedback from NRL guys who worked on NORM security • NORMS proposal is based on asymmetric crypto for downstream flow ⇒ it’s what the above I-D enables • NORMS proposal for upstream traffic is based on a (private?) TTP (Trusted Third Party), not necessarily a PKI certification authority. IMHO the proposal is not sufficiently mature at this level, but may contain interesting ideas to further develop ⇒ needs further discussion with the authors... INRIA Rhône-Alpes - V. Roca -
RMT security discussion I-D • no progress... • we (authors/group) clearly need to decide what we want to do with this doc... • our proposal: • restrict its scope a little bit... • it’s not a survey of all possible security aspects! • and decide that either we can come to IETF’72 with a consolidated document or we give up the work INRIA Rhône-Alpes - V. Roca -
FCAST Discussion INRIA Rhône-Alpes - V. Roca -
FCAST • what has been done so far? • initial I-D in March 2007 • a few discussions on the mailing list • a call for volunteers (still open...) • ...and that’s all for the moment • why another file transfer application? • it’s lighter (this is one of the goals) • it’s efficient with both ALC and NORM • it addresses more efficiently simple use-cases • it bypasses Nokia’s IPR claim on FLUTE INRIA Rhône-Alpes - V. Roca -
1- Is FCAST lighter? • is it really lighter? • yes if we remain in line with the 2000 proposal, i.e., append the meta-data to the file being sent in an aggregate object • what is supposed to be lighter? • the specifications... a little bit • the implementation and processing overhead... yes • because there’s no need to keep a database of the session’s objects and their meta-data • the meta-data of an object are processed immediately INRIA Rhône-Alpes - V. Roca -
Is FCAST lighter? (cont’) • because meta-data is processed only once, just on time • whereas “FDT Instances can be equal to, a subset of, a superset of, or complement any other FDT Instance” • ⇒ requires to process each and every FDT Instance • because the meta-data format/processing are simpler • in <draft-ietf-rmt-pi-alc-00>, there is no XML format but a simple list: <field>:<value>CR as for the HTTP object metainformation ⇒ no need for an XML parser • even with an XML format (other option) there would be fewer elements/attributes INRIA Rhône-Alpes - V. Roca -
Is FCAST lighter? (cont’) • because the sender does not have to care about the optimal transmission strategy for meta-data • whereas with FDT Instances in on-demand: what tx rate for FDT Instance? how often (even with static session)? INRIA Rhône-Alpes - V. Roca -
2- Is FCAST efficient with ALC and NORM? • I implemented FAST/ALC/NORM it in the past • the same application was running on top of both protocols without problem, with very minor per protocol adaptations • see http://planete-bcast.inrialpes.fr/ for more info and a (very old!) open-source implementation • FCAST/NORM is more efficient than FLUTE/NORM in on-demand mode • with FLUTE/NORM the FDT Instance is sent periodically • with FCAST/NORM meta-data is sent once, just on time INRIA Rhône-Alpes - V. Roca -
3- Is it more efficient with simple use-cases? • use-case considered: • each receiver is interested by all the objects sent in the LCT channel(s) he has joined • why is it more efficient? • FLUTE limited by FDT Inst/object interdependency • the object cannot be processed before processing the associated meta-data in an FDT Instance • FCAST is not • because meta-data is directly attached to the object INRIA Rhône-Alpes - V. Roca -
Is it more efficient... (cont’) • Rod’s suggestion to improve FLUTE’s behavior (email 01/24/08): • bundle meta-data (or subset) within an aggregate object • yes we can streamline the FDT Instance (and send it more often) and move a subset of meta-data in the aggregate objects... • ... but: • it adds one more layer of complexity: composite objects • it does not solve totally the interdependency problem since processing the FDT Instance is needed to know it’s a composite object INRIA Rhône-Alpes - V. Roca -
Is it more efficient... (cont’) • another good point for FCAST • meta-data benefits from the FEC encoding of the object • whereas an FDT Instance fits within a small # of packets which makes their protection difficult • ⇒ repetition is often the only solution to improve robustness with FLUTE INRIA Rhône-Alpes - V. Roca -
4- It bypasses Nokia’s IPR claim • if we stay inline with the initial proposal, as already said: • FCAST is described in the first ALC I-D: • http://tools.ietf.org/html/draft-ietf-rmt-pi-alc-00 • published on March 8th, 2000 • whereas Nokia’s IPR claim: • DATACAST FILE TRANSMISSION WITH META-DATA RETENTION • https://datatracker.ietf.org/public/ipr_detail_show.cgi?ipr_id=733 • priority date: January 31, 2003 INRIA Rhône-Alpes - V. Roca -
What’s next? • a final comment: • FLUTE is a versatile tool whereas FCAST is more specialized, with a more narrow scope. It’s important to keep this in mind while reading the previous slides... • if the group agrees... • Brian and myself propose a mature -01 version for IETF’72 • other volunteers are still welcome INRIA Rhône-Alpes - V. Roca -