On the Privacy Leakage of Coded Caching
Yu Wang, Alhussein A. Abouzeid · 2020
Coded caching, introduced by Maddah-Ali and Niesen, jointly designs content placement and delivery policies that achieve order-optimal transmission rate. We first prove that the coded caching scheme cannot protect the anonymity of network users when an eavesdropper in the network has access to the broadcast channel. Then, a three-stage attack is proposed to estimate users file requests by building an X-D table mapping between server transmissions X and users requests D. For a coded caching network having N files and K users, there are NKdifferent demand vectors. We show that the approximate average time complexity of the proposed attack is only a factor of two from the total number of possible demand vectors, while the best case (for the eavesdropper) time complexity is in the order of only NK. Once the X-D table is completely deduced by the eavesdropper, the users seize from having any privacy. Numerical simulations demonstrate that the analytical results for the time complexity of the attack match well with the simulation results.