70 likes | 305 Views
Wrap-up. Thanks!. Organization. Pay! Book of abstracts Sign your books in library Upload your abstracts Proceedings Executive summary Extended abstracts/drafts Case studies/benchmarks (upload!!). Tuesday’s Discussion. Protocols. Methodologies. Simulation Graph theoretical
E N D
Organization • Pay! • Book of abstracts • Sign your books in library • Upload your abstracts • Proceedings • Executive summary • Extended abstracts/drafts • Case studies/benchmarks (upload!!)
Tuesday’s Discussion Protocols Methodologies • Simulation • Graph theoretical • Graph rewriting/process alg. • Static analysis • Model checking • Quant. Model checking • Theorem proving • Control theory • API Authorization • Gossip • Chess wsn • Fraglet based (adaptive)
Some protocols we saw, benchmarks? • Chess WSN clock synchronization (Frits) • prob,. timed automata • Multipath routing (Nigel) • Graph optimisation problem • Fraglets • Alternating bit protocol using fraglets? (Frits, Christian) • DYMO (Bengt) • graph grammars, backwards reachability • AODV (Sebastian) • wireless process calculi, static analysis • ZigBee key establishment (Ender) • Static analysis • API Authorization (David B. Lee) • Automata minimisation (process calculi?) • Gossip (Rena) • Continous approximation of discrete behaviour, mean field
Random Quotes (Reminder) • Theorem proving is useless • Simulation just doesn’t work • ZigBee never supposed to be secure • Statistical model checking can help • Finite abstractions cannot help for gossip protocols • Model checking discovers corner cases • Analogy: scope extrusion – active networks (code within packets) • Process algebras with mobility could do, too • Theorem proving is expressive enough, but not out-of-the-box • Simulation can’t be trusted • Large scale gossip: hopeless state space • Process algebra tools are slow • Don’t trust models • In the limit, probabilistic systems are deterministic (mean-field, chemical master equation) • Static analysis useful to obtain models from implementation • MANETs just don’t work • Put some models on the table
Issues raised, discussions? • Model extraction/synthesis from implementation (using static analysis?) • Compilation of models to implementations • In what language do we write down our protocols? • Usable programming language for network modelling • For property specification • Process calculi/Markov chains/graph grammars/C programs? • Or our properties?