Possible Prevention. ( ) We designed the ACCEPTABLE( ) function in Section 2.3 so as to return false if C (or a cor- responding IP address) is being served already by another server instance upon receiving a new message C, m . Note that the ‘serving’ corresponds to the authentication session only. This was actually contrived for resisting the many-to-many guessing attack. For the purpose, a small hash table may be maintained by the server to track the currently served or blocked clients. The blocking policy should be considered carefully but flexibly. This resolution method may reduce the window of vulnerability notably but still leaves an issue about DoS (Denial of Service). Note that there is a recent literature considering DoS attacks on password-based protocols [9]. Aside from the danger of DoS attacks, a race condition and some bottleneck to the hash table are now only concerns while they could be negligible by careful consideration. The possible prevention methods might be considered both in the design and implementation phases. · ·
Possible Prevention. We designed the ACCEPTABLE( ) function in Section 2.3 so as to return false if C (or a corresponding IP address) is being served already by another server instance upon receiving a new message C, m . Note that the ‘serving’ corresponds to the authentication session only. This was actually contrived for resisting the many- to-many guessing attack. For the purpose, a small hash table may be maintained by the server to track the currently served or blocked clients. The blocking policy should be considered carefully but flexibly. This resolution method may reduce the window of vulnerability notably but still leaves an issue about DoS (Denial of Service). Note that there is a recent literature considering DoS attacks on password-based protocols [9]. Aside from the danger of DoS attacks, a race condition and some bottleneck to the hash table are now only concerns while they could be negligible by careful consideration. The possible prevention methods might be considered both in the design and implementation phases. In order to examine the reality of our attack, we implement a prototype of TP-AMP and launch the many-to-many guessing attack on it. We implement 4 This attack is negligible in the four-pass protocols since the server does not give sufficient information to the adversary forward and the client is usually not capable of listening to so many concurrent requests in the opposite case. Also, the best- known predecessors, EKE [5] and A-EKE [6], avoid this attack very impressively by not optimizing the protocol steps. · · · ⟨ ⟩ both client and server using CreateThread( ) functions and WinSock in Pen- tium IV 1.8GHz, 512MB, MS-Windows platforms. Simply a single run takes 297 milliseconds. We then drive multi-threaded clients to starts many simulta- neous authentication sessions. We summarize the results. Let δ = 5 (times) and ε = 3 (seconds), without correct ACCEPTABLE( ) function. Until we increase the number of adversarial client threads to 100, we could observe all requests are stably served and the same number of triples, m, µ, k1 , are gathered. When 150 threads are driven, about 20 requests are only declined, and about 130 triples are gathered. However, about 145 threads are blocked with correct ACCEPTABLE( ) function.
Possible Prevention. ( ) We designed the ACCEPTABLE( ) function in Section 2.3 so as to return false if C (or a corresponding IP address) is being served already by another server instance upon receiving a new message C, m . Note that the ‘serving’ corresponds to the authentication session only. This was actually contrived for resisting the many- to-many guessing attack. For the purpose, a small hash table may be maintained by the server to track the currently served or blocked clients. The blocking policy should be considered carefully but flexibly. This resolution method may reduce the window of vulnerability notably but still leaves an issue about DoS (Denial of Service). Note that there is a recent literature considering DoS attacks on password-based protocols [9]. Aside from the danger of DoS attacks, a race condition and some bottleneck to the hash table are now only concerns while they could be negligible by careful consideration. The possible prevention methods might be considered both in the design and implementation phases. In order to examine the reality of our attack, we implement a prototype of TP-AMP and launch the many-to-many guessing attack on it. We implement 4 This attack is negligible in the four-pass protocols since the server does not give sufficient information to the adversary forward and the client is usually not capable of listening to so many concurrent requests in the opposite case. Also, the best- known predecessors, EKE [5] and A-EKE [6], avoid this attack very impressively by not optimizing the protocol steps. · · ·
