Security Impact of Leakage Profiles: Threats and Countermeasures

Abstract

In Chap. 3, we discussed the security definitions of searchable encryption. In short, an SE scheme is said to achieve adaptively semantic security with a leakage function $\mathcal {L}(\cdot )$ if the information revealed during its operation is bounded by $\mathcal {L}(\cdot )$. Typically, allowed leakages include the size of encrypted data collection, whether a search token (trapdoor) has been repeated, which encrypted documents have been accessed, etc. In the extreme case, we can claim that a trivial plaintext solution is secure with $\mathcal {L}(\mathrm {DB}, w_1, \dots , w_t)=(\mathrm {DB}, w_1, \dots , w_t)$, i.e., it leaks all plaintext data, though this leakage function is meaningless since it obviously reveals confidential data to the untrusted server.