Mehmet Emin Dalkiliç

A Self Organizing Multihop Clustering Protocol for Wireless Sensor Networks

Wireless Sensor Network (WSN) self organization protocols are designed to build interconnection links among sensor nodes to create a manageable structure after the initial deployment. During the organization, clustering is a preferred way to achieve scalability. Because of the wireless transmission range and density limitations, connectivity and coverage per cluster can be improved with multihop links. [13] is one of the self organizing protocols in literature proposing “persistent”, a message efficient multihop clustering algorithm, together with a network decomposition method based on cluster initiator timers. In this paper, we have proposed a self organizing multihop clustering protocol which is also based on persistent algorithm. Different from the timer based approach in [13], our network decomposition is achieved by iterative local decisions and builds intercluster links while initiating new clusters. We have provided protocol details and investigated its performance by comparing simulation results with timer based approach in terms of amount of clusters created, message complexity and configuration time. Even though it has a slightly higher message complexity due to extra messages needed to generate intercluster links, the new protocol has a better clustering performance and operates well in low node densities. Configuration is much faster and the system scales well as the network size increases.

An Interactive Cryptanalysis Algorithm for the Vigenere Cipher

Though it dates back centuries, Vigenere Cipher is still a practical encryption method that can be edffciently used for many applications. We have designed and implemented an interactive cryptanalysis software based on the Kasiski test, and a novel use of the Index of Coincidence (IC) concept. Our results show that cryptanalysis is possible for very short text lengths where classical cryptanalysis methods fail. Furthermore, we have observed that our software which is designed to work on English based ciphertexts, can be successfully executed on ciphertexts based on other languages. Along the way, we also compute and report the IC values for Turkish and some other languages under a diffierent number of enciphering alphabets.

A simple shuffle-based stable in-place merge algorithm

Abstract Sorting is one of the most fundamental problems in computer science. Divide-and-conquer based sorting algorithms use successive merging to combine sorted arrays; therefore performance of merging is critical for these types of applications. The classic merge algorithm uses an extra O ( m + n ) memory for combining arrays of size m and n . Some improvements on merge algorithms include in-place methods which reduce or eliminate additional memory space, thus getting more practical value. We present a new in-place, stable and shuffle-based merge algorithm which is simple to understand and program. The algorithm starts applying perfect shuffle on two sorted arrays. Then, using the knowledge that odd and evenindexed numbers are sorted among themselves, comparisons are made and then misplaced elements are relocated by applying successive inverse perfect shuffle and swap operations on blocks. We have implemented and compared the new algorithm with the classical merge algorithm and the shuffle-based algorithm of Ellis and Markov [Comp. J. 1 (2000)]. Through experiments we have observed that the new algorithm exhibits linear run time behaviour and has dramatic performance improvement compared to the Ellis and Markov’s algorithm [1] .

The weak and the strong password preferences: a case study on turkish users

Considering the computer authentication, any password shall not only be private to its owner but also be chosen as not to be predicted easily by others. The passwords used in authentication processes of any critical system should be strong as not to be cracked easily. In this context, the strong password choice gets significance for the general system security. This study aims to reveal the proper and improper properties on password preferences via examining the real samples. The method used in this work is first to gather the real passwords in plaintext, then to crack the encrypted forms of them and finally to investigate statistical queries on those passwords in order to distinguish the common weak and strong characteristics. As the case study, the experiments are conducted on real passwords of Turkish users in an actively running system. The results of the experiments are categorized for weak and strong passwords. Moreover, the common tendencies on password choice are evaluated.

Cross layer implementation of key establishment protocols in wireless sensor networks

Security in Wireless Sensor Networks (WSN) can be achieved by establishing shared keys among the sensor nodes having possibility to communicate in the field. However, to achieve the optimum network operation after the deployment, there may be no need to establish keys for all the possible links. Because, links to be established are defined by the network configuration and as long as the network connectivity requirement is satisfied number of links to be secured can be limited accordingly. In this sense, key establishment and network configuration performances are related to each other and this cross relation should be considered when implementing security for WSN. In this paper, we have investigated the cross layer relations and performance variations of the selected pre-distribution and public key based key establishment protocols with the network configuration protocol proposed in [6]. We have proposed three methods as straight, reactive and proactive for selection of links to be secured and simulated the performance in terms of network connectivity, total energy cost and resilience to node capture attack. Simulation results indicate that total energy cost can be reduced by reducing the number of links to be secured without affecting the connectivity performance of the configuration protocol. Also we have figured out that the energy and resilience performance of the public key establishment is better than the key pre-distribution for the given network configuration parameters.

Experiments on a secure compression algorithm

A compression tool can provide security to some degree, if the algorithm itself is kept secret. Whenever the algorithm is known, as is the case for conventional compression, we cannot expect secrecy out of the tool. If security is needed as well, compression can be combined with encryption. To employ a system that would afford security while compressing, a novel technique (named as secure compression -SeCom) has been designed and implemented. We present the experimental results obtained by applying this scheme on texts from English and Turkish corpora. The results are promising in terms of compression quality and throughput, measured in bpc and MB/min, respectively.

Interconnect optimization in behavioral synthesis of ASICs

A new quadratic time global interconnect allocation algorithm for ASIC data path synthesis is presented. The algorithm is based on graph coloring and flipping commutative operation inputs and it has two phases: phase one sets an upper bound on number of multiplexer/bus inputs while phase two minimizes multiplexer/bus inputs as well as the number of two-input multiplexer equivalent of the interconnection network using interconnect merging techniques. The algorithm was applied to problems taken from current literature, and the results compare favorably to those for previous techniques.<<ETX>>

Homomorphic Minimum Bandwidth Repairing Codes

To store data reliably, a number of coding schemes including Exact-Minimum Bandwidth Regenerating codes (exact-MBR) and Homomorphic Self Repairing Codes (HSRC) exist. Exact-MBR offers minimum bandwidth usage whereas HSRC has low computational overhead in node repair. We propose a new hybrid scheme, Homomorphic Minimum Bandwidth Repairing Codes, derived from the above coding schemes. Our coding scheme provides two options for node repair operation. The first option offers to repair a node using minimum bandwidth and higher computational complexity while the second one repairs a node using fewer nodes, lower computational complexity and higher bandwidth. In addition, our scheme introduces a basic integrity checking mechanism.

Investigating the Effects of Recency and Size of Training Text on Author Recognition Problem

Prediction by partial match (PPM) is an effective tool to address the author recognition problem. In this study, we have successfully applied the trained PPM technique for author recognition on Turkish texts. Furthermore, we have investigated the effects of recency, as well as size of the training text on the performance of the PPM approach. Results show that, more recent and larger training texts help decrease the compression rate, which, in turn, leads to increased success in author recognition. Comparing the effects of the recency and the size of the training text, we see that the size factor plays a more dominant role on the performance.

On the Cryptographic Patterns and Frequencies in Turkish Language

Although Turkish is a significant language with over 60 million native speakers, its cryptographic characteristics are relatively unknown. In this paper, some language patterns and frequencies of Turkish (such as letter frequency profile, letter contact patterns, most frequent digrams, trigrams and words, common word beginnings and endings, vowel/consonant patterns, etc.) relevant to information security, cryptography and plaintext recognition applications are presented and discussed. The data is collected from a large Turkish corpus and the usage of the data is illustrated through cryptanalysis of a mono-alphabetic substitution cipher. A new vowel identification method is developed using a distinct pattern of Turkish--(almost) non-existence of double consonants at word boundaries.