Hidden Credential Retrieval (HCR) protocols are designed for access credentials management where users who remember short passwords can retrieve his/her various credentials (access keys and tokens) with the help of a remote storage server over insecure networks (e.g., the Internet). In this paper, we revisit two HCR protocols, both of which are based on blind signature schemes: one (we call it B-HCR) was proposed in ASIACCS 2009 and the other (we call it MRS-HCR) was in WISA 2010. In particular, we show that the B-HCR protocol is insecure against an outside attacker who impersonates server S. Specifically, the attacker can find out the user's password pw with off-line dictionary attacks by eavesdropping the communications between the user and a third-party online service provider. Also, we show that the MRS-HCR protocol does not work correctly itself. In other words, user U can not retrieve the plaintext Msg (i.e., credentials) even if he/she has a knowledge of the password.

An augmented PAKE (Password-Authenticated Key Exchange) protocol provides password-only authentication in the presence of an attacker, establishment of session keys between the involving parties, and extra protection for server compromise (i.e., exposure of password verification data). Among many augmented PAKE protocols, AMP variants (AMP2 [16] and AMP+ [15]) have been standardized in IEEE 1363.2 [9] and ISO/IEC 11770-4 [10]. In this paper, we thoroughly investigate APKAS-AMP (based on AMP2 [16]) and KAM3 (based on AMP+ [15]) which require several validity checks on the values, received and computed by the parties, when using a secure prime. After showing some attacks on APKAS-AMP and KAM3, we suggest new sanity checks that are clear and sufficient to prevent an attacker from doing these attacks.

An anonymous password-authenticated key exchange (PAKE) protocol is designed to provide both password-only authentication and client anonymity against a semi-honest server, who honestly follows the protocol. In INDOCRYPT2008, Yang and Zhang [26] proposed a new anonymous PAKE (NAPAKE) protocol and its threshold (D-NAPAKE) which they claimed to be secure against insider attacks. In this paper, we first show that the D-NAPAKE protocol [26] is completely insecure against insider attacks unlike their claim. Specifically, only one legitimate client can freely impersonate any subgroup of clients (the threshold t > 1) to the server. After giving a security model that captures insider attacks, we propose a threshold anonymous PAKE (called, TAP+) protocol which provides security against insider attacks. Moreover, we prove that the TAP+ protocol has semantic security of session keys against active attacks as well as insider attacks under the computational Diffie-Hellman problem, and provides client anonymity against a semi-honest server, who honestly follows the protocol. Finally, several discussions are followed: 1) We also show another threshold anonymous PAKE protocol by applying our RATIONALE to the non-threshold anonymous PAKE (VEAP) protocol [23]; and 2) We give the efficiency comparison, security consideration and implementation issue of the TAP+ protocol.

An augmented PAKE (Password-Authenticated Key Exchange) protocol is said to be secure against server-compromise impersonation attacks if an attacker who obtained password verification data from a server cannot impersonate a client without performing off-line dictionary attacks on the password verification data. There are two augmented PAKE protocols where the first one [12] was proposed in the IEEE Communications Letters and the second one [15] was submitted to the IEEE P1363.2 standard working group [9]. In this paper, we show that these two augmented PAKE protocols [12], [15] (claimed to be secure) are actually insecure against server-compromise impersonation attacks. More specifically, we present generic server-compromise impersonation attacks on these augmented PAKE protocols [12],[15].

Secure channels can be realized by an authenticated key exchange (AKE) protocol that generates authenticated session keys between the involving parties. In, Shin et al., proposed a new kind of AKE (RSA-AKE) protocol whose goal is to provide high efficiency and security against leakage of stored secrets as much as possible. Let us consider more powerful attacks where an adversary completely controls the communications and the stored secrets (the latter is denoted by "replacement" attacks). In this paper, we first show that the RSA-AKE protocol is no longer secure against such an adversary. The main contributions of this paper are as follows: (1) we propose an RSA-based leakage-resilient AKE (RSA-AKE2) protocol that is secure against active attacks as well as replacement attacks; (2) we prove that the RSA-AKE2 protocol is secure against replacement attacks based on the number theory results; (3) we show that it is provably secure in the random oracle model, by showing the reduction to the RSA one-wayness, under an extended model that covers active attacks and replacement attacks; (4) in terms of efficiency, the RSA-AKE2 protocol is comparable to in the sense that the client needs to compute only one modular multiplication with pre-computation; and (5) we also discuss about extensions of the RSA-AKE2 protocol for several security properties (i.e., synchronization of stored secrets, privacy of client and solution to server compromise-impersonation attacks).

An anonymous password-authenticated key exchange (anonymous PAKE) protocol is designed to provide both password-only authentication and user anonymity against a semi-honest server, who follows the protocol honestly. Very recently, Yang and Zhang have proposed a new anonymous PAKE (NAPAKE) protocol that is claimed efficient compared to the previous constructions. In this paper, we propose a very-efficient anonymous PAKE (called, VEAP) protocol that provides the most efficiency among their kinds in terms of computation and communication costs. The VEAP protocol guarantees semantic security of session keys in the random oracle model under the chosen target CDH problem, and unconditional user anonymity against a semi-honest server. If the pre-computation is allowed, both the user and the server are required to compute only one modular exponentiation, respectively. Surprisingly, this is the same computation cost of the well-known Diffie-Hellman protocol that does not provide authentication at all. In addition, we extend the VEAP protocol in two ways: the first is designed to reduce the communication costs of the VEAP protocol and the second shows that stripping off anonymity parts from the VEAP protocol results in a new PAKE protocol.