Depth-robust graphs and their cumulative memory complexity

J.F. Alwen, J. Blocki, K.Z. Pietrzak, in:, J.-S. Coron, J. Buus Nielsen (Eds.), Springer, 2017, pp. 3–32.


Conference Paper | Published | English
Editor
;
Department
Series Title
LNCS
Abstract
Data-independent Memory Hard Functions (iMHFS) are finding a growing number of applications in security; especially in the domain of password hashing. An important property of a concrete iMHF is specified by fixing a directed acyclic graph (DAG) Gn on n nodes. The quality of that iMHF is then captured by the following two pebbling complexities of Gn: – The parallel cumulative pebbling complexity Π∥cc(Gn) must be as high as possible (to ensure that the amortized cost of computing the function on dedicated hardware is dominated by the cost of memory). – The sequential space-time pebbling complexity Πst(Gn) should be as close as possible to Π∥cc(Gn) (to ensure that using many cores in parallel and amortizing over many instances does not give much of an advantage). In this paper we construct a family of DAGs with best possible parameters in an asymptotic sense, i.e., where Π∥cc(Gn) = Ω(n2/ log(n)) (which matches a known upper bound) and Πst(Gn) is within a constant factor of Π∥cc(Gn). Our analysis relies on a new connection between the pebbling complexity of a DAG and its depth-robustness (DR) – a well studied combinatorial property. We show that high DR is sufficient for high Π∥cc. Alwen and Blocki (CRYPTO’16) showed that high DR is necessary and so, together, these results fully characterize DAGs with high Π∥cc in terms of DR. Complementing these results, we provide new upper and lower bounds on the Π∥cc of several important candidate iMHFs from the literature. We give the first lower bounds on the memory hardness of the Catena and Balloon Hashing functions in a parallel model of computation and we give the first lower bounds of any kind for (a version) of Argon2i. Finally we describe a new class of pebbling attacks improving on those of Alwen and Blocki (CRYPTO’16). By instantiating these attacks we upperbound the Π∥cc of the Password Hashing Competition winner Argon2i and one of the Balloon Hashing functions by O (n1.71). We also show an upper bound of O(n1.625) for the Catena functions and the two remaining Balloon Hashing functions.
Publishing Year
Date Published
2017-04-01
Volume
10212
Page
3 - 32
Conference
EUROCRYPT: Theory and Applications of Cryptographic Techniques
Conference Location
Paris, France
Conference Date
2017-04-30 – 2017-05-04
IST-REx-ID

Cite this

Alwen JF, Blocki J, Pietrzak KZ. Depth-robust graphs and their cumulative memory complexity. In: Coron J-S, Buus Nielsen J, eds. Vol 10212. Springer; 2017:3-32. doi:10.1007/978-3-319-56617-7_1
Alwen, J. F., Blocki, J., & Pietrzak, K. Z. (2017). Depth-robust graphs and their cumulative memory complexity. In J.-S. Coron & J. Buus Nielsen (Eds.) (Vol. 10212, pp. 3–32). Presented at the EUROCRYPT: Theory and Applications of Cryptographic Techniques, Paris, France: Springer. https://doi.org/10.1007/978-3-319-56617-7_1
Alwen, Joel F, Jeremiah Blocki, and Krzysztof Z Pietrzak. “Depth-Robust Graphs and Their Cumulative Memory Complexity.” edited by Jean-Sébastien Coron and Jesper Buus Nielsen, 10212:3–32. Springer, 2017. https://doi.org/10.1007/978-3-319-56617-7_1.
J. F. Alwen, J. Blocki, and K. Z. Pietrzak, “Depth-robust graphs and their cumulative memory complexity,” presented at the EUROCRYPT: Theory and Applications of Cryptographic Techniques, Paris, France, 2017, vol. 10212, pp. 3–32.
Alwen JF, Blocki J, Pietrzak KZ. 2017. Depth-robust graphs and their cumulative memory complexity. EUROCRYPT: Theory and Applications of Cryptographic Techniques, LNCS, vol. 10212. 3–32.
Alwen, Joel F., et al. Depth-Robust Graphs and Their Cumulative Memory Complexity. Edited by Jean-Sébastien Coron and Jesper Buus Nielsen, vol. 10212, Springer, 2017, pp. 3–32, doi:10.1007/978-3-319-56617-7_1.

Link(s) to Main File(s)
Access Level
OA Open Access

Export

Marked Publications

Open Data IST Research Explorer

Search this title in

Google Scholar
ISBN Search