Artificial Immune System (AIS)-based evolutionary algorithms combine rules and randomness to solve optimization and classification problems. Due to their capability in identifying self and non self samples, they have also gained attention in intrusion detection systems. In this paper, we propose a real-time AIS-based anomoly detection algorithm for intrusion detection. The most important features of the proposed method are its high detection rate, low false alarm, low computational complexity, and real-time response to the incoming samples. We compare our proposed method with several well-known anomaly detection algorithms on various datasets. We demonstrate that the proposed method performs the best among others in terms of false alarm, detection rate and time response.
Malaria is a serious disease with a high developing in the world. It still major public health problem. In 2008, 109 countries were declared as endemic to the disease, 243 million malaria cases were reported causing nearly a million of death, primarily of children under 5 year; nearly 3000 children die every day in Africa. In Brazil, Amazonia is the area that presents the highest number of cases and consequently having many cases of morbidity and mortality, especially when we deal with the Malaria caused by the Plasmodium falciparum. The parasite in the human organism attacks liver cells and erythrocytes. Plasmodium falciparum is the deadliest protozoan because attacks a high number of erythrocytes at any evolutionary phase, making the person susceptible to others disease. The human immune system (HIS) operates in the invasion of this parasite (innate system) and during its development in the organism (adaptive system), it is responsible for finding and neutralizing cells that are infected to prevent their proliferation, avoiding the disease or the death of the patient. The complex biology of Plasmodium falciparum and its interaction with the HIS are the motivation of this work. Therefore, with mathematical models and computer simulations which have been growing considerably in the medical science area, this work presents a model to contribute quickly and trustfully for this interaction. The mathematical modeling used is structured by non-linear ordinary differential equations of first-order that describes the action of the HIS to eliminate the protozoa. We have performed several simulations to investigate the behavior of the protozoa front of the human organism without the action of the HIS, and the results were compared to the behavior with the action of the HIS.
An negative selection algorithm is presented for intrusion detection tasks for systems with arbitrary diversity. This algorithm uses two types of agents, detectors and presenters. Presenters present information to detectors; detectors are selected to engage in a maximally frustrated dynamics when presenters present data from a reference state. We show that if presenters present information that has never been available during the selection stage, then presenters engage in a less frustrated dynamics and their abnormal presentation can be detected. The performance of our algorithm is independent of the dimension of the space, i.e., the length of information presented by presenters, and hence does not suffer from the dimensionality curse accompanying current methods.
Clustering is one of the most well known activities in scientific investigation and the object of research in many disciplines, ranging from Statistics to Computer Science. In this beautiful area, one of the most difficult challenges is the model selection problem, i.e., the identification of the correct number of clusters in a dataset. In the last decade, a few novel techniques for model selection, representing a sharp departure from previous ones in statistics, have been proposed and gained prominence for microarray data analysis. Among those, the stability-based methods are the most robust and best performing in terms of prediction, but the slowest in terms of time. Unfortunately, this fascinating and classic area of statistics as model selection, with important practical applications, has received very little attention in terms of algorithmic design and engineering. In this paper, in order to partially fill this gap, we highlight: (A) the first general algorithmic paradigm for stability-based methods for model selection; (B) a novel algorithmic paradigm for the class of stability-based methods for cluster validity, i.e., methods assessing how statistically significant is a given clustering solution; (C) a general algorithmic paradigm that describes heuristic and very effective speed-ups known in the Literature for stability-based model selection methods.
Data mining is the process of discovering patterns from large data sets. One of the branches of data mining is Associative Classification (AC). AC mining is a promising approach that uses association rules discovery techniques to construct association classifiers. However, traditional AC algorithms typically search for all possible association rules to find a representative subset of those rules. Since the search space of such rules may grow exponentially as the support threshold decreases, the rules discovery process can be computationally expensive. One effective way to tackle this problem is to directly find a set of high-stakes association rules that potentially builds a highly accurate classifier. This paper introduces AC-CS, a novel AC algorithm, inspired by the clonal selection of the immune system. The algorithm proceeds in an evolutionary fashion to populate only rules that are likely to yield good classification accuracy. Empirical results on several real datasets show that the approach generates dramatically less rules than traditional AC algorithms. Hence, the proposed approach is indeed significantly more efficient than traditional AC algorithms while achieving a competitive accuracy.
Some common systems modelling and simulation approaches for immune problems are Monte Carlo simulations, system dynamics, discrete-event simulation and agent-based simulation. These methods, however, are still not widely adopted in immunology research. In addition, to our knowledge, there is few research on the processes for the development of simulation models for the immune system. Hence, for this work, we have two contributions to knowledge. The first one is to show the importance of systems simulation to help immunological research and to draw the attention of simulation developers to this research field. The second contribution is the introduction of a quick guide containing the main steps for modelling and simulation in immunology, together with challenges that occur during the model development. Further, this paper introduces an example of a simulation problem, where we test our guidelines.
The vertebrate immune system is a complex distributed system capable of learning to tolerate the organisms' tissues, to assimilate a diverse commensal microflora, and to mount specific responses to invading pathogens. These intricate functions are performed almost flawlessly by a self-organised collective of cells. The robust mechanisms of distributed control in the immune system could potentially be deployed to design multiagent systems. However, the essence of the immune system is clonal expansion by cell proliferation, which is difficult to envisage in most artificial multiagent systems. In this paper, we investigate under which conditions proliferation can be approximated by recruitment in fixed-sized agent populations. Our study is the first step towards bringing many of the desirable properties of the adaptive immune system to systems made of agents which are incapable of self-replication. We adopt the crossregulation model of the adaptive immune system. We develop ordinary differential equation models of proliferation-based and recruitment-based systems, and we compare the predictions of these analytical models with results obtained by a stochastic simulation. Our results define the operational parameter regime wherein growth by recruitment retains all the properties a cell proliferation model. We conclude that rich immunological behaviour can be fully recapitulated in sufficiently large multiagent systems based on growth by recruitment.
Summary: We propose the in vivo/vitro use of prokaryotic adaptive immune systems for distributed learning. In the coming years synthetic biologists will learn to control, program, and modify such systems. We design an enhancement to CRISPR-Cas immune systems and demonstrate the learning potential of the modified system by showing it can approximate solutions to a computationally hard problem. To our knowledge this is the first proposed use of CRISPR-Cas systems for computational purposes.
Link failure and unreachable nodes due to interference from external devices are common problems in WSNs. These interferences can be a major inhibitor to node performance and network stability. In order to tolerate these failures, we propose an immune-inspired self healing system where an individual node can detect degradations in network performance, perform diagnostic tests, and provide automated immediate response to recover the network to a stable state. We evaluate and compare the performance of our approach with other routing protocols on a testbed environment using TelosB hardware motes.
The negative selection algorithm is one of the oldest immune-inspired classification algorithms and was originally intended for anomaly detection tasks in computer security. After initial enthusiasm, performance problems with the algorithm lead many researchers to conclude that negative selection is not a competitive anomaly detection technique. However, in recent years, theoretical work has lead to substantially more efficient negative selection algorithms. Here, we report the results of the first evaluation of negative selection with r-chunk and r-contiguous detectors that employs these novel algorithms. On a collection of 14 datasets from real-world sources, we compare negative selection with r-chunk and r-contiguous detectors against techniques based on kernels, finite state automata, and n-gram frequencies, and find that negative selection performs competitively, yielding a slightly better average performance than all other techniques investigated. Because this study represents, to our knowledge, the most comprehensive one of string-based negative selection to date, the widely held view that negative selection is not a competitive anomaly detection technique may be inaccurate.
Algorithms based on graphs and networks offer great potential for modeling of physiological processes. In this paper bipartite graphs called Petri Nets (PN) are presented and their utility in modeling of the immune system is highlighted. We present our improved PN model of the immune system. The coupling between fever and autism is studied using this model. It is shown that fever may shift the level of IL-1, IL-6 cytokines in autistic subjects closer to standard physiological values. This is in accordance with a recent observation that fever improves behavior of autistic children, reported in medical literature.
Artificial Immune System (AIS) achieved some success in malware detection with its distributed, diverse and adaptive characteristics. However, in recent years, malware is evolving quickly in respect of stealth and complexity. This trend has brought a great challenge for AIS, especially when spyware emerged. To solve this problem, natural killer cells (NKs) which can lure latent viruses to expose themselves are introduced to AIS in this paper. We hope their counterparts can enhance the anti-latent capability of AIS by enticement strategy and collaboration with other AIS algorithms. Preliminary results show that artificial NKs can discover tiny abnormalities caused by novel spyware, and then release proper bait (called induction cytokines) to trigger the spyware’s actions which will expose itself to further detection by AIS.
This paper begins by stating that the underlying concepts of signals and antigen used by the Dendritic Cell Algorithm are too abstract and arbitrary to be of use in real world applications as they stand. To address this, these concepts are more explicitly defined within a specific application area, namely that of data stream analysis. These new definitions are based around the outputs of the Change Point Detecting Subspace Tracker (CD-ST), a recently developed algorithm for detecting key change points across multiple data streams. Preliminary results demonstrate the utility of this new definition for antigen. The paper concludes by laying the theoretical groundwork for a novel anomaly detection framework for use in data streaming applications. The underlying methodology is to perform anomaly detection via the detection and classification of key change points that occur across the multiple data streams monitored.
Metabolic P systems are a modeling framework for metabolic, regulatory and signaling processes. The synthesis of flux regulation functions from time series of substance concentrations is a key task for reverse-engineering biological systems by MP systems. In this paper we present some important improvements to a technique based on genetic algorithms and multiple linear regression for the synthesis of regulation functions. An accurate analysis of generated functions, for the case study of the mitotic oscillator in early amphibian embryos, shows that some knowledge about the regulation mechanisms of biological processes can be inferred from experimental data using this methodology.
The Dendritic Cell Algorithm (DCA) is an immune inspired algorithm based on the behavior of dendritic cells. The performance of DCA depends on the selected features and their categorization to their specific signal types, during pre-processing. For feature selection, DCA applies the Principal Component Analysis (PCA). Nevertheless, PCA does not guarantee that the selected first principal components will be the most adequate for classification. Furthermore, the DCA categorization process is based on the PCA attributes' ranking in terms on variability. However, this categorization process could not be considered as a coherent assignment procedure. Thus, the aim of this paper is to develop a new DCA feature selection and categorization method based on Rough Set Theory (RST). In this model, the selection and the categorization processes are based on the RST CORE and REDUCT concepts. Results show that applying RST, instead of PCA, to DCA is more convenient for data pre-processing yielding much better performance in terms of accuracy.
Computing a longest common subsequence of a number of strings is a classical combinatorial optimisation problem with many applications in computer science and bioinformatics. It is a hard problem in the general case so that the use of heuristics is motivated. Evolutionary algorithms have been reported to be successful heuristics in practice but a theoretical analysis has proven that a large class of evolutionary algorithms using mutation and crossover fail to solve and even approximate the problem efficiently. This was done using hard instances. We reconsider the very same hard instances and prove that the B-cell algorithm outperforms these evolutionary algorithms by far. The advantage stems from the use of contiguous hypermutations. The result is another demonstration that relatively simple artificial immune systems can excel over more complex evolutionary algorithms in the domain of optimisation.
The presented work proposes a new approach for anomaly detection. This approach is based on changes in a population of evolving agents under stress. If conditions are appropriate, changes in the population (modeled by the bioindicators) are representative of the alterations to the environment. This approach, based on an ecological view, improves functionally traditional approaches to the detection of anomalies. To verify this assertion, experiments based on Network Intrussion Detection Systems are presented. The results are compared with the behaviour of other bioinspired approaches and machine learning techniques.
Artificial Immune Systems (AIS) [1] include algorithms and systems that use the human immune system as inspiration. The human immune system is a robust, decentralised, error tolerant and adaptive system. Such properties are highly desirable for the development of novel computer systems, but also - we would like to say - for characterizing complex systems and for contributing to the growth of complexity science.
Face recognition algorithms often have to filter out the disturbances of some conditional factors such as facial pose, illumination, and expression (PIE). So an increasing number of researchers have been figuring out the best discrimi-nant transformation in the feature space of faces to improve the recognition performance. They have also proposed novel feature-matching algorithms to minimize the PIE effects. For example, Chen et al. designed a nearest feature space (NFS) embedding algorithm that outperformed the other algorithms for face recognition. By searching the most similar sample with immune learning, in this paper, a novel algorithm is proposed to filter out the disturbances of PIE for face recognition. The adaptive adjustment for filtering out the disturbance of PIE is designed with immune memory to maximize the success possibility for recognizing the faces. The clonal selection frame is used to search the most similar samples to the target face, and the selected antibodies are memorized as the candidates for the best solution or the second optimal solution. The proposed approach is evaluated on several benchmark databases and is compared with the NFS embedding algorithm. The experimental results show that the proposed approach outperforms the NFS embedding algorithm.
This paper presents a new artificial immune system algorithm for solving multi-objective optimization problems, based on the clonal selection principle and the hypervolume contribution. The main aim of this work is to investigate the performance of this class of algorithm with respect to approaches which are representative of the state-of-the-art in multi-objective optimization using metaheuristics. The results obtained by our proposed approach, called multi-objective artificial immune system based on hypervolume (MOAIS-HV) are compared with respect to those of the NSGA-II. Our preliminary results indicate that our proposed approach is very competitive, and can be a viable choice for solving multi-objective optimization problems.
