Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Naming in Content-Oriented Architectures 1 Data publishing RWI select produce own Data Name certify Key 2 “Basic bindings” The ICN paper argued that RWI, Name, and Key should be bound together “If not, then it is impossible to identify the principal of the object” Name RWI “If not, then anyone could claim to be the principal associated with the data” “If not, then the receiver doesn’t know which public key to use to verify the provenance of the data” Key 3 Binding between RW and cyberspace Real world RWI Who should take care of the bindings associated with RWI? Network or users? Cyberspace Data Name Key 4 “NDN bindings” in the ICN paper RWI The paper argued “The purpose of human-readability is to establish an intrinsic binding between the name and the RWI” Name The paper argued “The binding (between the name and key) can be established using an external authority.” Key 5 The true story about NDN security • Security of NDN does not rely on human-readability – human-readability just provides contexts for consumers to understand and remember names • NDN secure the binding between name and data – decision on which key is legitimate to certify the binding is left to consumers – consumers define their own trust management policies – the key can be derived from consumers’ own policies Data Name certify Key 6 SCN bindings RWI The paper said “the binding between the RWI and key must be established by an external authority” “Name” Key Label The paper argued that “The use of cryptographic hash … provides the binding between the name and the key, enabling the receiver to check the key” Key 7 Unresolved security issues in SCN • It is more difficult to make SCN name correct – SCN name consist of two parts: producers’ key and data label. – both parts must be correct – NDN name is data name only, no requirement on producer’s key • The ICN paper suggested several methods of getting the right key – search engine based trust – social network based trust – the ICN paper did not elaborate how to achieve the goal using these methods 8 DoS • In content-oriented network, DoS means “consumer cannot get the requested data” • The ICN paper argued that – DoS can be prevented if routers can reject false data – “content-oriented routing must be done on a name-key basis, not just on the name” – “this makes the key an essential part of the name” 9 Different views • SCN trusts the network as a whole to return the correct data – in fact, given a name, network may not always return the correct data – consumer has no choice when receiving the false data • NDN assumes it is possible to receive false data from the network – intelligent forwarding is used to avoid “bad” routers or hosts in the network – when receiving false data, consumer can notify routers to try some other paths 10 Scalability of SCN routing • In order to fetch data C, concatenation of the form A.B.C is used • A, B, and C are all selfcertifying names • Upon receiving a request for A.B.C, look up A, B, and C in routing table • Take the deepest match Router 1 A.B.C name A B F out-if 1 2 3 ✔ out-if 1 3 4 ✔ Router 2 A.B.C name B C F 11 back to hierarchical name • Concatenation in the ICN paper is actually hierarchical name – "The semantics of such concatenations (A.B.C) are that when following routing entries for A, you will eventually find one for B; and that when following routing entries for B, you will eventually find an entry for C.” in Section 3.1 • The only difference is the component of a concatenation is flat name • The ICN paper also admitted that the label part in SCN name could also be hierarchical – “This hybrid choice merely requires that the labels in the P:L format for self-certifying names be human readable and hierarchical.” in Section 2.2 Denial of Service 12 Flexibility of SCN • The ICN paper was focused on the flexibility of trust model • The paper argued that NDN has to rely on PKI, because – network must be able to verify data – without key as a part of name, decentralized trust is impossible 13 Flexibility is not just about trust model • NDN is more flexible than SCN in trust model – NDN, from very beginning, allows decentralized trust • consumers decide if the signing key can be trusted – given a binding between name and data • NDN allows multiple signers • SCN allows only one signer, which is bound to the name • NDN is more flexible than SCN in naming – NDN does not restrict the format of names – SCN name can be represented as NDN name • two-component NDN name: producer’s key and data name • concatenation in SCN is also NDN name as well – several pairs of producers’ key and data name – NDN name cannot be represented as SCN name • SCN name requires key which does not exist in NDN name 14 General process of data fetching • NDN – given a data name – send an interest to network – get the data with signature and signer info – consumer determines whether the signer can be trusted – if so, accept the data – if not, notify router to try some other paths • SCN – given a data label – consumer determines who is the producer – consumer gets producer’s key – construct a SCN name • Hash(key) + label – send request – assume returned data is always correct 15 NDN vs. SCN • The goals of NDN1 are – security: NDN is as secure as SCN • NDN allows consumers to decide which key can be trusted • NDN use intelligent forwarding to prevent against DoS – usability: NDN name is more useful • Human-readable name is easier to use than flat name • NDN name stays with data, while SCN name has to changes due to routine key replacement – flexibility: NDN is more flexible than SCN in general • More flexible trust model • More flexible naming mechanism • The ICN paper also mentioned scalability – SCN resolved back to hierarchical name which has been used by NDN 1. Diana Smetters and Van Jacobson, “Securing Network Content”, NDN TR 16