Thiago S. Guzella

Review: A review of machine learning approaches to Spam filtering

In this paper, we present a comprehensive review of recent developments in the application of machine learning algorithms to Spam filtering, focusing on both textual- and image-based approaches. Instead of considering Spam filtering as a standard classification problem, we highlight the importance of considering specific characteristics of the problem, especially concept drift, in designing new filters. Two particularly important aspects not widely recognized in the literature are discussed: the difficulties in updating a classifier based on the bag-of-words representation and a major difference between two early naive Bayes models. Overall, we conclude that while important advancements have been made in the last years, several aspects remain to be explored, especially under more realistic evaluation settings.

Identification of SPAM messages using an approach inspired on the immune system.

In this paper, an immune-inspired model, named innate and adaptive artificial immune system (IA-AIS) is proposed and applied to the problem of identification of unsolicited bulk e-mail messages (SPAM). It integrates entities analogous to macrophages, B and T lymphocytes, modeling both the innate and the adaptive immune systems. An implementation of the algorithm was capable of identifying more than 99% of legitimate or SPAM messages in particular parameter configurations. It was compared to an optimized version of the naive Bayes classifier, which has been attained extremely high correct classification rates. It has been concluded that IA-AIS has a greater ability to identify SPAM messages, although the identification of legitimate messages is not as high as that of the implemented naive Bayes classifier.

Towards a novel immune inspired approach to temporal anomaly detection

In this conceptual paper, we report on studies and initial definitions of an immune-inspired approach to temporal anomaly detection problems, where there is a strict temporal ordering on the data, such as intrusion detection and fault detection. The inspiration for the development of this approach comes from the sophisticated mechanisms involved in T-cell based recognition, such as tuning of activation thresholds, receptor down-regulation, among others. Despite relying on low affinity and highly degenerate interactions, the recognition of foreign patterns by T cells is both highly sensitive and specific. Through a proper understanding of some of these mechanisms, this work aims at developing an efficient computational model using some of these concepts.

A novel immune inspired approach to fault detection

This paper presents a novel immune inspired algorithm, named DERA (Dynamic Effector Regulatory Algorithm), aimed at fault detection and other anomaly detection problems. It integrates immunological research results, pointing out the importance of a special breed of cells (regulatory T cells) in the control of various aspects of the immune system, and includes a mechanism for signalling between cells. Preliminary results of the application of the proposed model to the DAMADICS fault detection dataset are presented, indicating that the proposed approach can attain good results when properly tuned.

Regulatory T cells: inspiration for artificial immune systems

In this conceptual paper, some features of regulatory T cells are described. These cells have been receiving an increasing attention in Immunological research, due to their importance in several aspects of the immune system. As will be argued, these cells constitute an important source of inspiration for developing Artificial Immune Systems, computational tools that attempt to capture some of the characteristics of the natural immune system. It is expected that the incorporation of these cells in some immune inspired algorithms may not only lead to more biologically plausible models, but also to algorithms that can achieve better results in real-life problems.

Proposta de um modelo de classificação de padrões baseado no sistema imune: uma aplicação para a identificação de SPAM

Resumo—Neste artigo, é proposto um modelo de classificação de padrões baseado no sistema imune. Esse modelo, denominado IA-AIS (Innate and Adaptive Artificial Immune System), é baseado na Teoria de Seleção Clonal e em componentes dos sistemas imune inato e adaptativo, tais como macrófagos, linfócitos B e T. O modelo proposto foi implementado para classificação de SPAM (mensagens de e-mail comerciais não solicitadas, enviadas em massa e automaticamente). Em testes realizados, foram obtidas taxas de acerto superiores a 98% na distinção entre SPAM e mensagens legı́timas. Para efeito de comparação, os resultados foram confrontados com os obtidos a partir do Modelo Bayesiano de classificação também implementado.

On the Relevance of Cellular Signaling Pathways for Immune-Inspired Algorithms

In this conceptual paper, we discuss the relevance of cellularsignaling pathways for immune-inspired algorithms. With complexdynamics, the mapping of environment stimuli to cellular responsesis highlighted as a decision making capability. When consideringapplications which could benefit from these dynamics, thepossibility of incorporating these pathways can be an interestingway to combine more biologically-plausible algorithms and improvedperformance. The structure of the NF-κB (NuclearFactor κB) and MAP (Mitogen-activated protein)kinases pathways, and the pathways involved in signaling byToll-like receptors, are presented. As an example, we then considerhow these pathways could be incorporated in the Dendritic CellAlgorithm.