Luiz G. A. Martins

Some Investigations About Synchronization and Density Classification Tasks in One-dimensional and Two-dimensional Cellular Automata Rule Spaces

The study of computational aspects of cellular automata (CA) is a recurrent theme being that the investigation of specific tasks to be solved by CA rules a common and widely-known approach. We investigated two of the most-studied computational tasks: synchronization (ST) and density classification (DCT). Different specifications of CA rule space were analyzed for both tasks: one-dimensional rules with radius 1 and 2, and two-dimensional rules with von Neumann and Moore neighborhoods. We also analyzed different lattice sizes when trying to execute these tasks. Several evolutionary experiments were performed to characterize ST and DCT on these different scenarios. Some interesting results have been occurred from these experiments as the adequacy of the tasks to be solved in two-dimensional spaces instead of 1D even using rules with the same length and the dependency to the parity of the lattice size related to good rules for DCT in 1D and 2D spaces.

Secret key specification for a variable-length cryptographic cellular automata model

Reverse algorithm was previously evaluated as encryption method concluding that its simple adoption is unviable, since it does not assurance the pre-image existence. Variable-Length Encryption Method (VLE) was proposed where a alternative algorithm with extra bits is adopted when pre-image computation is not possible. If an adequate secret key is used with VLE it is expected that the final ciphertext length is close to plaintext size. Several CA static parameters were calculated for a set formed by all radius 2 right-toggle rules. A database was generated associating rules performance in VLE ciphering with its parameters. A genetic algorithm-based data mining was performed to discover an adequate key specification based on CA parameters. Using such specification, ciphertext length is short, encryption process returns high entropy and VLE has a good protection against differential cryptanalysis.

Exhaustive evaluation of radius 2 toggle rules for a variable-length cryptographic cellular automata-based model

A cellular automata (CA) model in cryptography is investigated. A previous work analyzed the usage of reverse algorithm for pre-image computation as an encryption method. The main conclusion was that the simple adoption of such method is not viable, since it does not have 100% of guarantee of pre-image existence. A new approach was proposed that uses extra bits when the pre-image computation is not possible. It is expected that in practice few failures happens and the ciphertext size will be close to the plaintext. Encryption always succeeds and the final length of the ciphertext is not fixed. We better investigate the secret key specification by using a more representative set formed by all radius 2 right-toggle rules, totalizing 65536 rules. An exhaustive analysis of this rule space has shown that using adequate specification the method has a good protection against differential cryptanalysis and a small increase in ciphertext length.

Architecture of a remote impedance-based structural health monitoring system for aircraft applications

The essence of structural health monitoring (SHM) is to develop systems based on nondestructive inspection (NDI) technologies for continuous monitoring, inspection and detection of structural damages. A new architecture of a remote SHM system based on Electromechanical Impedance (EMI) measures is described in the present contribution. The proposed environment is employed to automatically monitor the structural integrity of aircrafts and is composed by sensor networks, a signal conditioning system, a data acquisition hardware and a data processing system. The obtained results allow the accomplishment of structural condition-based maintenance strategies, in opposite to those based only on the usage time of the equipment. This approach increases the operational capacity of the structure without compromising the security of the flights. As the environment continually checks for the first signs of damage, possibly reducing or eliminating scheduled aircraft inspections, it could significantly decrease maintenance and repair expenses. Furthermore, the usage of this system allows the creation of a historical database of the aircrafts structural integrity, making possible the incremental development of a Damage Prognosis System (DPS). This work presents the proposed architecture and a set of experiments that were conducted in a representative aircraft structure (aircraft window) to demonstrate the effectiveness of the proposed system.

MO-Miner: A Data Mining Tool Based on Multi-Objective Genetic Algorithms

Data mining (DM) is the process to extract previously unknown and implicit information from large databases. Several techniques have been used to discover such kind of knowledge; most of them derived from machine learning and statistics. The majority of these approaches focus on the discovery of accurate knowledge. However, this knowledge should be useless if it does not offer some kind of surprisingness to the final user. Based on this idea, some investigations were started recently with the aim of extracting accurate and interesting information from datasets. In this sense, data mining can be faced as a multiobjective problem. Different machine learning approaches have been employed to perform data mining tasks. Most of them are based on evolutionary methods, like genetic algorithms (Goldberg, 1989). Another advantage of genetic algorithms is that they can be adapted to treat multi-objective problems in a Pareto sense. Various multi-objective genetic algorithms have been proposed in the literature, like the method known as Non-dominated Sorting Genetic Algorithms (NSGA) (Srinivas & Deb, 1994). The task performed in a data mining process depending on what kind of knowledge someone needs to extract. The main types of tasks performed by DM algorithms are classification, association, clustering, regression, sequences analysis, summarizing and dependence modelling. Classification task searches for the knowledge able to predict the value of a previously defined goal attribute based on other attributes. This knowledge is often represented by IF-THEN rules and it is the most investigated data mining task. Dependence modelling task can be seen as a generalization of classification. It also aims to discover rules able to predict the goal attribute value, from values of the prediction attributes. However, in dependence modelling, there are more than one goal attribute. Initially, a small set of goal attributes is specified, whose prediction is considered interesting. These attributes may occur in the consequent or in the antecedent parts of the rule. The others attributes occur only in the antecedent. A multi-objective evolutionary data mining environment named MO-miner was implemented based on the family of algorithms called non-dominated sorting genetic algorithms (NSGA). The two desirable properties of the rules being mined accuracy and interestingness are simultaneously manipulated. MO-miner keeps the metrics related to O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m

An improved robot path planning model using cellular automata

Bio-inspired techniques have been successfully applied to the path-planning problem. Amongst those techniques, Cellular Automata (CA) have been seen a potential alternative due to its decentralized structure and low computational cost. In this work, an improved CA model is implemented and evaluated both in simulation and real environments using the e-puck robot. The objective was to construct a collision-free path plan from the robot initial position to the target position by applying the refined CA model and environment pre-processed images captured during its navigation. The simulations and real experiments show promising results on the model performance for a single robot.

Decision Tree-Based Ensembles To Specify Secret Keys For Cellular Automata Cryptographic Model

A cryptographic method based on cellular automata (CA) was previously proposed which employs transition rules as secret keys. However, some rules belonging to the possible key space present undesirable behaviors that must be avoided. In a previous work, it was investigated the secret key specification for this cryptography model associating rules performance in ciphering with CA static parameters. A genetic algorithm-based data mining was performed to discover adequate key specification and it was employed to filter the set of all possible radius 2 CA rules. It was able to discover good secret key specifications. However, such filter provokes a significant decay in the number of good keys, while still keeping some underperforming rules. Adequate secret key specifications are investigated here using decision tree ensembles: bootstrap aggregating (bagging), boosting and random forest. The new filters are compared to the previous ones. By applying the new methodology, it was possible to find filters able to eliminate almost all underperforming rules and keeping a higher number of adequate secrete keys.

Algoritmos Genéticos Multi-Objetivos Na Mineração De Regras Interessantes E Precisas

 This work evaluates the use of multi-objective genetic algorithms (MOGA) in the mining of accurate and interesting rules. For this, a program was implemented based on the MOGA technique called nondominated sorting genetic algorithm (NSGA), which was applied in the database Zoo of public domain. The results of our experiments had been compared with those generated by a standard genetic algorithm in order to identify the benefits related to the multi-objective approach. Keyword Multi-objective genetic algorithms, rules mining, data mining. Resumo Este trabalho avalia o uso de algoritmos genéticos multi-objetivos (AGMO) no processo de mineração de regras precisas e interessantes. Implementou-se um programa baseado na técnica de AGMO conhecida como nondominated sorting genetic algorithm, o qual foi aplicado na base de dados de domínio público Zoo. Os resultados foram comparados com aqueles gerados por um algoritmo genético padrão, a fim de identificar as melhorias obtidas pela adoção de uma abordagem multi-objetivos. Palavras-chave Algoritmos genéticos multi-objetivos, extração de regras, mineração de dados.

Adaptive strategies applied to evolutionary search for 2D DCT cellular automata rules

Cellular automata (CA) are able to perform complex computations through local interactions. The investigation of how CA computations are carried out can be made by the usage of CA rules to solve specific tasks. The well-known problem called density classification task (DCT) is investigated, with focus on its two-dimensional version. Evolutionary algorithms have been widely used in the search for DCT rules. A sample of lattices with Gaussian distribution is commonly used to evaluate rule quality. However, uniform lattices are easier to classify, allowing an initial selective pressure needed to start the convergence. A comparative evaluation of three adaptive strategies is presented here: they start using easy lattices to classify and as effective rules are being obtained the difficult level is progressively increased toward the target evaluation. Several experiments were performed to evaluate the strategies efficiency and new rules were found, which outperform the best ones published.