Jun Kurihara

The traditional Domain Name System (DNS) lacks fundamental features of security and privacy in its design. As concerns of privacy increased on the Internet, security and privacy enhancements of DNS have been actively investigated and deployed. Specially for user's privacy in DNS queries, several relay-based anonymization schemes have been recently introduced, however, they are vulnerable to the collusion of a relay with a full-service resolver, i.e., identities of users cannot be hidden to the resolver. This paper introduces a new concept of a multiple-relay-based DNS for user anonymity in DNS queries, called the mutualized oblivious DNS ($\mu$ODNS), by extending the concept of existing relay-based schemes. The $\mu$ODNS introduces a small and reasonable assumption that each user has at least one trusted/dedicated relay in a network and mutually shares the dedicated one with others. The user just sets the dedicated one as his next-hop, first relay, conveying his queries to the resolver, and randomly chooses its $0$ or more subsequent relays shared by other entities. Under this small assumption, the user's identity is concealed to a target resolver in the $\mu$ODNS even if a certain (unknown) subset of relays collude with the resolver. That is, in $\mu$ODNS, users can preserve their privacy and anonymity just by paying a small cost of sharing its resource. Moreover, we present a PoC implementation of $\mu$ODNS that is publicly available on the Internet. We also show that by measurement of round-trip-time for queries, and our PoC implementation of $\mu$ODNS achieves the performance comparable to existing relay-based schemes.

In this paper, we first introduce Shamir’s construction of the (k; n)-threshold scheme as a typical linear secret-sharing scheme and explain the construction of the (k; l; n)-threshold ramp scheme proposed by Yamamoto and Blakley-Meadows as its extension. Then, by generalizing threshold ramp schemes with a linear code C1 and its subcode C2, we represent a linear secret-sharing scheme in terms of C1 and C2. As examples, we represent Shamir’s (k; n)-threshold scheme and the (k; l; n)-threshold ramp scheme of Yamamoto and Blakley-Meadows using linear codes. Furthermore, we show that in linear secret-sharing schemes, the maximum amount of information leakage of a secret message and their strong security are characterized by the relative generalized Hamming weights (RGHW’s) of C1 and C2 when every share is an element of a finite field.