Dr. CHETHAN CHANDRA S BASAVARADDI

Dagstuhl Reports, 2017

This report documents the program and the topics discussed of the 4-day Dagstuhl Seminar 17181 "Theory and Applications of Hashing", which took place May 1-5, 2017. Four long and eighteen short talks covered a wide and diverse range of topics within the theme of the workshop. The program left sufficient space for informal discussions among the 40 participants.

2012

Vulnerability of dedicated hash functions to various attacks has made the task of designing hash function much more challenging. This provides us a strong motivation to design a new cryptographic hash function viz. HF-hash. This is a hash function, whose compression function is designed by using first 32 polynomials of HFE Challenge-1 (6) with 64 variables by forcing remaining 16 variables as zero. HFhash gives 256 bits message digest and is as efficient as SHA-256. It is secure against the differential attack proposed by Chabaud and Joux in (5) as well as by Wang et. al. in (23) applied to SHA-0 and SHA-1. 1

Content based information system in which content is used for information retrieval. User sends query to Content Based Information Retrieval System (CBIR). If content is sensitive and user don't want to disclose this query to Content Based Information Retrieval system then there is need of privacy preservation. In previous system, that is content-based multimedia information retrieval contains browsing and also search paradigms, learning, semantic queries, new features and different media types, high performance indexing, and evaluation methods. The major challenge in the MIR system is semantic search with significance on the discovery of concepts in media with difficult backgrounds. The other system gives efficient algorithms for the exact nearest neighbour problem. Robust Sparse Hashing (RSH) is Nearest Neighbour (NN) retrieval. This approach is boosted by the victory of dictionary understanding for sparse coding. Vector Quantization (VQ) is magnificent quantization method from signal processing that permits the modeling of feasibility density function by the allocation of prototype vectors. Fast Library Appropriate Nearest Neighbour is a library for accomplishing fast appropriate Nearest Neighbour (NN) search in high dimensional spaces. A Discrete Cosine Transform (DCT) conveys a finite sequence of data points in terms of a sum of cosine function moving at different frequency. In current content based information retrieval system, one problem is that both client and server sides are loaded with the same tasks. To solve this problem, the common task is assigned to the third party in such a way that privacy should also be preserved. This framework is tested using large image database. It provides two layers of protection. First, robust hash value are used as queries to secure original content. Second, in future to decrease the ambiguity of the server the client can choose to omit some bits in hash values, two different robust hash algorithm are used. One is random projection and other is discrete wavelet transform. This improves the performance of retrieval.

ACM SIGCSE Bulletin, 2002

Hashing is a singularly important technique for building efficient data structures. Unfortunately, the topic has been poorly understood historically, and recent developments in the practice of hashing have not yet found their way into textbooks.

1996

, the Brazilian community i n Waterloo, and my host family Ted, Carlene, Lawrence, Matthew and Mickey Goddard, for all the pleasant memories I left behind. Finally, I w ant to thank my family for their unconditional love during these years. My warmest feelings go to my wife Graciela and to my daughter Manuelita for providing meaning in my life. Almost one year ago, Graciela and Manuelita returned to Uruguay, and two months later I visited them for one short week. Half an hour before I left to return to Canada, Manuelita grabbed my hand and began a conversation that marked one of the special moments in my life. In many w ays, this conversation guided me in this my nal year of research. My memory of it was so strong that one day I w as inspired to write a short story in Spanish about it. I feel that Manuelita deserves a privileged place in this thesis and so, at the beginning of each c hapter I quote some fragments of this story, starting with its rst sentence immediately prior to Chapter 1 and ending with its last prior to Chapter 6. vi To m y d a u g h ter Manuelita and the moon, the sources of my inspiration and love. vii Bibliography x 2 CHAPTER 1. INTRODUCTION 1.1 Introduction The idea of hashing seems to have been originated by H. P. Luhn, in an internal IBM memorandum in January 1953 46]. The rst major paper published in the area is the classic article by P eterson 69]. In this work, Peterson de nes open addressing in general, and gives empirical statistics about linear probing hashing. He also notes the degradation in performance when records are deleted. Moreover, he acknowledges that the open addressing idea was devised in 1954 by A.L. Samuel, G.M. Amdahl, and E. Boehme. A good early survey of the area is the paper by W. B u c hholz 12]. Nevertheless, as noted by K n uth 46], the word \hashing" to identify this technique appeared for the rst time in the literature in the survey of Morris 63], although it had been in common usage for several years. In that paper he introduced the idea of random probing (with secondary clustering). Linear probing is the simplest collision resolution for open addressing. It works reasonably well for tables that are not too full, but as the load factor increases, its performance deteriorates rapidly. The longer a contiguous sequence of key grows, the more likely collisions with this sequence will occur when new keys are inserted. Furthermore, one insertion may coalesce two long clusters. This phenomenon is called primary clustering. The main application of linear probing is to retrieve information in secondary storage devices when the load factor is not too high, as rst proposed by P eterson 69]. It was also proposed by Larson as a method to handle over ow records in linear hashing schemes 56, 5 7 ]. One reason for the use of linear probing is that it preserves locality of reference between successive probes, thus avoiding long seeks 55]. The rst published analysis of linear probing for buckets of size 1, was done by Konheim and Weiss 52]. However, this algorithm was rst analyzed by K n uth in 1962 46, 82], who stated that this analysis had a strong in uence in the structure of his series \The Art of Computer Programming". A di erent approach to the analysis of this hashing scheme, based on the application of ballot theorems, was presented by Mendelson and Yechiali 61]. P ug and Kessler 70] study the case in which the keys are nonuniformly distributed. They do an asymptotic analysis for the case in which the size of the table tends to in nity while the load factor is constant. Pittel 71], also presents an asymptotic analysis of the probable largest cost of a successful search. Finally, Aldous 3] studies the case when the access probabilities of the keys are not uniform. Operating primarily in the context of double hashing, several authors 10, 4 , 3 6 ] observed that a collision could be resolved in favor of any of the keys involved, and used this additional degree of freedom to decrease the expected search time in the table. We obtain the standard schemes by letting the incoming key probe its next location. Celis et al. 16, 1 7 ] w ere the rst to observe that collisions could be resolved having variance reduction as a goal. They de ned the Robin Hood heuristic, in which e a c h collision occurring on each insertion is resolved in favor of the key that is farthest away from its home location. Later, Poblete and Munro 73] de ned the last-come-rst-served heuristic,