Semi-supervised anomaly detection in dynamic communication networks

OHODIN is an online extension for data streams of the kNN-based ODIN anomaly detection approach. It provides a detection-threshold heuristic that is based on extreme value theory. In contrast to sophisticated anomaly and novelty detection approaches the decision-making process of ODIN is interpretable by humans, making it interesting for certain applications. However, it is limited in terms of the underlying detection method. In this article, we present an extension of the OHODIN to further detection techniques to reinforce OHODIN capability of online data streams anomaly detection. We introduce the algorithm modifications and an experimental evaluation with competing state-of-the-art anomaly detection approaches.

Detecting frauds in computing platforms involves identifying malicious user activity sessions. Recently, deep learning models have been employed to design fraud detection approaches. Effective training of these deep learning models requires a large amount of well-annotated sessions. However, due to the cost of expert annotation, many organizations rely on heuristics to perform automated annotation, which leads to the noisy label learning problem. It is well known that the performance of deep learning models can easily degrade because of noisy or inaccurate labels. To tackle this challenge, we propose a supervised Contrastive Learning based Fraud Detection (CLFD) framework, which is designed to operate in the noisy label setting. CLFD employs an effective label corrector for correcting noisy labels and which is specifically designed for the fraud detection task. Then, by employing the corrected labels, it trains a fraud detector through supervised contrastive learning, and derives separable representations. We empirically evaluate our CLFD framework and other state-of-the-art baselines on benchmark datasets. Our CLFD framework demonstrates superior performance over state-of-the-art baselines.

There are many factors in mucosal secretions that contribute to innate immunity and the 'first line of defence' at mucosal surfaces. Few studies, however, have investigated the effects of exercise on many of these 'defence' factors. The aim of the present study was to determine the acute effects of prolonged exercise on salivary levels of selected antimicrobial peptides (AMP) that have not yet been studied in response to exercise (HNP1-3 and LL-37) in addition to immunoglobulin A (IgA). A secondary objective was to assess the effects of exercise on saliva antibacterial capacity. Twelve active men exercised on a cycle ergometer for 2.5 h at approximately 60% of maximal oxygen uptake. Unstimulated whole saliva samples were obtained before and after exercise. There was a significant decrease (P < 0.05) in salivary IgA:osmolality ratio, following exercise, but IgA concentration and secretion rate were unaltered. Salivary HNP1-3 and LL-37 concentrations (P < 0.01 and P < 0.05, respectively), concentration:osmolality ratios (P < 0.01) and secretion rates (P < 0.01) all increased following exercise. Salivary antibacterial capacity (against E. coli) did not change. The increased concentration of AMPs in saliva may confer some benefit to the 'first line of defence' and could result from synergistic compensation within the mucosal immune system and/or airway inflammation and epithelial damage. Further study is required to determine the significance of such changes on the overall 'defence' capacity of saliva and how this influences the overall risk for infection.

Today, the proportion of elderly and senile people is steadily growing throughout the world. The most important factors in the first line of immune defense of the mucous membranes of the upper respiratory tract are β-defensins, which are a group of secretory proteins with antimicrobial activity. The aim of this study was to study the expression of genes for antimicrobial peptides β-defensins and the composition of the microbiome of the mucous membranes of the upper respiratory tract in elderly people and long-livers with various aging phenotypes.The main study group included 67 centenarians and 49 elderly people, who were further divided into two subgroups depending on the course of aging (pathological and successful aging). Nucleic acids were isolated from nasopharyngeal scrapings and the expression levels of the DEFB1 and DEFB4 genes were determined using real-time polymerase chain reaction. The composition of the microbiota in nasopharyngeal swabs was determined by MALDI-TOF mass spectrometry.In analyzing the expression of the DEFB1 gene in elderly people and centenarians with successful and pathological aging phenotypes, no difference was revealed between the groups. Expression of the DEFB4 gene was increased in centenarians with pathological aging compared to centenarians with successful aging and in the elderly group. Excessive production of antimicrobial peptides is dual in nature; on the one hand, they provide the first line of defense against microorganisms, and on the other, they are cytotoxic to their own cells. An increase in the expression of the DEFB4 gene during aging may be due to an increase in the number of pathogen-associated molecular patterns, which can be one’s own microbiota and/or components of microbial metabolism. Analysis of the microbiota composition showed an increase in biodiversity in individuals with a successful aging phenotype compared to a pathological phenotype. Particular attention is paid to Staphylococcus spp., the species composition of which depends on the aging phenotype. In the pathological aging group, the frequency of St. aureus colonization is significantly higher than in the successful aging group.Thus, overexpression of the DEFB4 gene and changes in the composition of the microbiota of the mucous membranes of the upper respiratory tract may be one of the mechanisms explaining the increased susceptibility to infections in various aging phenotypes.

Authentication systems: A literature review and classification

By comparing natural immunity to Aspergillus fumigatus (AF) in vivo with the action of human or mouse phagocytes against AF in vitro, we delineated two sequential lines of defense against AF. The first line of defense was formed by macrophages and directed against spores. Macrophages prevented germination and killed spores in vitro and rapidly eradicated conidia in vivo, even in neutropenic and athymic mice. The second was the neutrophilic granulocyte (PMN), which protected against the hyphal form of AF. Human and mouse PMN killed mycelia in vitro. Normal, but not neutropenic mice, stopped hyphal growth, and eradicated mycelia. Either line of defense acting alone protected mice from high challenge doses. Natural immunity collapsed only when both the reticuloendothelial system and PMN were impaired. These findings are in keeping with the clinical observation that high doses of cortisone and neutropenia are the main risk factors for invasive aspergillosis. Cortisone inhibited the conidiacidal activity of mouse macrophages in vivo and of human or mouse mononuclear phagocytes in vitro. Cortisone damaged this first line of defense directly and not through the influence of T lymphocytes or other systems modifying macrophage function as shown in athymic mice and in vitro. In addition, daily high doses of cortisone in mice reduced the mobilization of PMN so that the second line of defense was also impaired. Thus, cortisone can break down natural resistance on its own. Myelosuppression rendered mice susceptible only when the first line of defense was overpowered by high challenge doses, by activated spores that cannot be killed by macrophages, or by cortisone suppression of the conidiacidal activity of macrophages. The host, thus, can call upon two independent phagocytic cell lines that form graded defense systems against aspergillus. These lines of defense function in the absence of a specific immune response, which seems superfluous in the control and elimination of this fungus.

Citation: 'first line of defence' in the IUPAC Compendium of Chemical Terminology, 5th ed.; International Union of Pure and Applied Chemistry; 2025. Online version 5.0.0, 2025. 10.1351/goldbook.13151 • License: The IUPAC Gold Book is licensed under Creative Commons Attribution-ShareAlike CC BY-SA 4.0 International for individual terms. Requests for commercial usage of the compendium should be directed to IUPAC.

BGP, the protocol responsible for interdomain rout- ing across the global internet, continues to face critical vul- nerabilities, particularly in the form of routing anomalies and prefix hijacking attacks. Despite the existence of various anomaly detection systems, most current solutions suffer from poor scalability and require frequent, resource-intensive retraining to accommodate the continuous addition of new Autonomous Systems (ASes) and the emergence of novel attack patterns. In this work, we propose a multi-tiered anomaly detection framework that allows network operators and service providers to select different tiers of security based on their operational priorities and available computational resources. This will allow the network to respond quickly, in case of events where new ASes are added to the global network or new attacks are discovered. This tier-based approach ensures that lightweight, rapidly deployable detection mechanisms are available as a first line of defense, while more comprehensive, resource-heavy models continue training in the background.

Droughts pose significant challenges for accurate monitoring due to their complex spatiotemporal characteristics. Data-driven machine learning (ML) models have shown promise in detecting extreme events when enough well-annotated data is available. However, droughts do not have a unique and precise definition, which leads to noise in human-annotated events and presents an imperfect learning scenario for deep learning models. This article introduces a 3-D convolutional neural network (CNN) designed to address the complex task of drought detection, considering spatiotemporal dependencies and learning with noisy and inaccurate labels. Motivated by the shortcomings of traditional drought indices, we leverage supervised learning with labeled events from multiple sources, capturing the shared conceptual space among diverse definitions of drought. In addition, we employ several strategies to mitigate the negative effect of noisy labels (NLs) during training, including a novel label correction (LC) method that relies on model outputs, enhancing the robustness and performance of the detection model. Our model significantly outperforms state-of-the-art drought indices when detecting events in Europe between 2003 and 2015, achieving an AUROC of 72.28%, an AUPRC of 7.67%, and an ECE of 16.20%. When applying the proposed LC method, these performances improve by +5%, +15%, and +59%, respectively. Both the proposed model and the robust learning methodology aim to advance drought detection by providing a comprehensive solution to label noise and conceptual variability.

Experimental modulation of cellular glutathione levels has been used to explore the role of glutathione in cadmium toxicity. Mice treated with buthionine sulfoximine [an effective irreversible inhibitor of gamma-glutamylcysteine synthetase (EC 6.3.2.2) that decreases cellular levels of glutathione markedly] were sensitized to the toxic effects of CdCl2. Mice pretreated with a sublethal dose of Cd2+ to induce metallothionein synthesis were not sensitized to Cd2+ by buthionine sulfoximine. Mice sensitized to Cd2+ by buthionine sulfoximine were protected against a lethal dose of Cd2+ by glutathione mono isopropyl ester (L-gamma-glutamyl-L-cysteinylglycylisopropyl ester), but not by glutathione. These results are in accord with studies that showed that glutathione mono esters (in contrast to glutathione) are efficiently transported into cells and converted intracellularly to glutathione. The findings indicate that intracellular glutathione functions in protection against Cd2+ toxicity, and that this tripeptide provides a first line of defense against Cd2+ before induction of metallothionein synthesis occurs. The experimental approach used here in which cellular levels of glutathione are decreased or increased seems applicable to investigation of other types of metal toxicity and of other glutathione-dependent biological phenomena.

Most interactions between plants and fungi or Oomycetes do not result in a successful infection but are thwarted by the host (Thordal-Christensen, 2003; Lipka et al., 2008). In many cases, the obvious visible first line of defence is the papilla, a cell wall apposition (CWA) laid down by the host at the site of attempted penetration by the fungus or Oomycete on the inside of the host cell wall. CWA are observed in many plant species but are best studied in the cereal–powdery mildew interactions where the CWA comprises the inner papilla surrounded by an outer halo. Histological and chemical analyses demonstrate the complexity of the papillae, which comprise callose, protein, and various phenolic compounds and inorganic compounds, especially opal silica and, at least transiently, reactive oxygen species (see Zeyen et al., 2002, for a review). It is clear that papillae play a significant role, since penetration is often arrested by papillae, but as to which components of papillae are important for successful defence against a particular species pathogen is unclear. In other words, the demonstration of the complexity of papillae is not a demonstration of their role in disease resistance and presents the question as to why they are so complex. What is the role of the individual components of the papilla? There are two approaches which use the tools of forward and reverse genetics to determine the role of individual components of papillae which will answer this question, at least in part. The regulation of the formation of papillae has been studied genetically in both barley and Arabidopsis. Mutants have been obtained in barley exhibiting resistance, and indeed the mlo mutant confers race-non-specific resistance in both species (Consonni et al., 2006). Other Arabidopsis mutations contribute to the assembly of papillae (Collins et al., 2003, 2007). The physiological role of the Arabidopsis gene products is uncertain in some cases (Lipka et al., 2008). Thus the forward genetics approach has favoured the identification of regulatory, like mlo, and mechanistic components, like Pen1, but has not yielded mutations in the phenolic biosynthetic pathways associated with papilla formation. The nature of the first line of defence in cereals is still something of a mystery despite many years of study. We know much but there is still much to learn. Wei and colleagues have previously developed Triticum monococcum L. (Einkorn, a diploid wheat) as an excellent model system to use as an alternative to barley. In this issue, the study by Bhuiyan et al. (2009) uses a forward genetic approach to study the role of papilla components in Einkorn wheat infected with the powdery mildew fungus Blumeria graminis. The result is compelling evidence that lignification of papillae plays an important role in defence against penetration by this fungus. Lignification is, in essence, the generation of wood. The process can be considered to comprise two phases, firstly, the biosynthesis of monolignols, the building bricks, and, secondly, the assembly and polymerization of these bricks in the papilla. This paper represents an in-depth study of the first phase, which is illustrated in Fig. 1 of Bhuiyan et al. (2009). The approach taken has been to screen an expression sequence tag (EST) library prepared from the epidermis of T. monococcum infected with B. graminis f. sp. tritici (Bgt) (Liu et al., 2005; Bhuiyan et al., 2007) which is predicted to be enriched for first line defences against Blumeria graminis (Wei et al., 1998; Collinge et al., 2002). This yielded 13 cDNAs representing eight genes encoding enzymes involved in monolignol biosynthesis, according to their bioinformatical analysis. That six of these transcripts exhibit nearly synchronized accumulation patterns, as revealed by Northern blotting of entire leaves and epidermal expression, as revealed by RT PCR, is an excellent indication of their role in papilla formation. Proof of the importance of monolignol biosynthesis comes from the reverse genetics of transient gene silencing of four of the biosynthetic genes and an additive effect was obtained by combining three of the genes pair-wise. Thus, silencing gave a statistically significant increase in the efficiency of penetration. In the control, roughly 40% of penetration attempts were successful by the host powdery mildew fungus Bgt and increased 4-fold in the non-host powdery mildew fungus, Bg. f.sp. hordei, which infects barley. The double silenced constructs gave over 70% successful penetration rates. The genes included TmPAL, which encodes the first enzyme in the phenylpropanoid pathway, and can be predicted to have other roles in defence in grasses. For example, flavonoid biosynthesis has been seen to be induced in barley in response to B. graminis (Christensen et al., 1998a, b). The combination of TmPAL with genes encoding the two final biosynthetic enzymes of

In this study, we present a novel model called ADA-NAF (Anomaly Detection Autoencoder with the Neural Attention Forest) for semi-supervised anomaly detection that uniquely integrates the Neural Attention Forest (NAF) architecture which has been developed to combine a random forest classifier with a neural network computing attention weights to aggregate decision tree predictions. The key idea behind ADA-NAF is the incorporation of NAF into an autoencoder structure, where it implements functions of a compressor as well as a reconstructor of input vectors. Our approach introduces several technical advances. First, a proposed end-to-end training methodology over normal data minimizes the reconstruction errors while learning and optimizing neural attention weights to focus on hidden features. Second, a novel encoding mechanism leverages NAF’s hierarchical structure to capture complex data patterns. Third, an adaptive anomaly scoring framework combines the reconstruction errors with the attention-based feature importance. Through extensive experimentation across diverse datasets, ADA-NAF demonstrates superior performance compared to state-of-the-art methods. The model shows particular strength in handling high-dimensional data and capturing subtle anomalies that traditional methods often do not detect. Our results validate the ADA-NAF’s effectiveness and versatility as a robust solution for real-world anomaly detection challenges with promising applications in cybersecurity, industrial monitoring, and healthcare diagnostics. This work advances the field by introducing a novel architecture that combines the interpretability of attention mechanisms with the powerful feature learning capabilities of autoencoders.

Collaborative representation-based detector (CRD), as the most representative anomaly detection method, has been widely applied in the field of hyperspectral anomaly detection (HAD). However, the sliding dual window of the original CRD introduces high computational complexity. Moreover, most HAD models only consider a single spectral or spatial feature of the hyperspectral image (HSI), which is unhelpful for improving detection accuracy. To solve these problems, in terms of speed and accuracy, we propose a novel anomaly detection approach, named Random Collective Representation-based Detector with Multiple Feature (RCRDMF). This method includes the following steps. This method first extract the different features include spectral feature, Gabor feature, extended multiattribute profile (EMAP) feature, and extended morphological profile (EMP) feature matrix from the HSI image, which enables us to improve the accuracy of HAD by combining the multiple spectral and spatial features. The ensemble and random collaborative representation detector (ERCRD) method is then applied, which can improve the anomaly detection speed. Finally, an adaptive weight approach is proposed to calculate the weight for each feature. Experimental results on six hyperspectral datasets demonstrate that the proposed approach has the superiority over accuracy and speed.

The ubiquity of the Internet connection to desktops has been both a boon to business as well as a cause for concern for the security of digital assets that may be unknowingly exposed. Firewalls have been the most commonly deployed solution to secure corporate assets against intrusions, but firewalls are vulnerable to errors in configuration, ambiguous security policies, data-driven attacks through allowed services, and insider attacks. The failure of firewalls to adequately protect digital assets from computer-based attacks has been a boon to commercial intrusion detection tools. Two general approaches to detecting computer security intrusions in real time are misuse detection and anomaly detection. Misuse detection attempts to detect known attacks against computer systems. Anomaly detection uses knowledge of users' normal behavior to detect attempted attacks. The primary advantage of anomaly detection over misuse detection methods is the ability to detect novel and unknown intrusions. This paper presents a study in employing neural networks to detect the existence of anomalous and unknown intrusions against a software system using the anomaly detection approach.

Recently, the major challenging attack is the side channel attack for both symmetric and asymmetric algorithm. Because in both cases, the secret key has to be sent out via a safe secured channel. The advanced encryption standard (AES) is mentioned as the best standard encryption algorithm by the US government, but there is one major attack on the AES that is the brute force attack. So, how to rule out these kinds of attack and how it is implemented effectively in hardware are being discussed in this paper. The honey encryption scheme is an effective algorithm, and it is also being discussed. In this paper, the detailed description of the AES and the methods to improve its speed and how it is integrated with the honey technique have been discussed. The different kinds of attacks and the various solutions are discussed for both of the algorithms. A survey of papers which has used these algorithms for various reasons with different perspective has been given. The honey algorithm creates the fake message for the wrong try of the key puncher. Thus, this is considered as the best algorithm, but it is also has some disadvantages that is also focussed in our survey. Copyright © 2017 John Wiley &amp; Sons, Ltd.