In this paper we make two observations on Rabin's probabilistic primality test. The first is a provocative reason why Rabin's test is so good. It turned out that a single iteration has a nonnegligible probability of failing _only_ on composite numbers that can actually be split in expected polynomial time. Therefore, factoring would be easy if Rabin's test systematically failed with a 25% probability on each composite integer (which, of course, it does not). The second observation is more fundamental because is it _not_ restricted to primality testing: it has consequences for the entire field of probabilistic algorithms. The failure probability when using a probabilistic algorithm for the purpose of testing some property is compared with that when using it for the purpose of obtaining a random element hopefully having this property. More specifically, we investigate the question of how reliable Rabin's test is when used to _generate_ a random integer that is probably prime, rather than to _test_ a specific integer for primality. Key words: factorization, false witnesses, primality testing, probabilistic algorithms, Rabin's test.

Abstract. Recently, Damg˚ard, Landrock and Pomerance described a procedure in which a k-bit odd number is chosen at random and subjected to t random strong probable prime tests. If the chosen number passes all t tests, then the procedure will return that number; otherwise, another k-bit odd integer is selected and then tested. The procedure ends when a number that passes all t tests is found. Let pk,t denote the probability that such a number is composite. The authors above have shown that pk,t ≤ 4 −t when k ≥ 51 and t ≥ 1. In this paper we will show that this is in fact valid for all k ≥ 2 and t ≥ 1. 1.

Abstract—In this paper, we uncover the Bloom paradox in Bloom filters: sometimes, it is better to disregard the query results of Bloom filters, and in fact not to even query them, thus making them useless. We first analyze conditions under which the Bloom paradox occurs in a Bloom filter, and demonstrate that it depends on the a priori probability that a given element belongs to the represented set. We show that the Bloom paradox also applies to Counting Bloom Filters (CBFs), and depends on the product of the hashed counters of each element. In addition, both for Bloom filters and CBFs, we suggest improved architectures that deal with the Bloom paradox. We also provide fundamental memory lower bounds required to support element queries with limited false-positive and false-negative rates. Last, using simulations, we verify our theoretical results, and show that our improved schemes can lead to a significant improvement in the performance of Bloom filters and CBFs. A. The Bloom Paradox