HAL-Université de Bretagne Occidentale
Not a member yet
75442 research outputs found
Sort by
Diffusion-based Adversarial Purification for Intrusion Detection
No full textInternational audienceThe escalating sophistication of cyberattacks has encouraged the integration of machine learning techniques in intrusion detection systems, but the rise of adversarial examples presents a significant challenge. These crafted perturbations mislead ML models, enabling attackers to evade detection or trigger false alerts. In response, adversarial purification has emerged as a compelling solution, particularly with diffusion models showing promising results. However, their purification potential remains unexplored in the context of intrusion detection. This paper demonstrates the effectiveness of diffusion models in purifying adversarial examples in network intrusion detection. Through a comprehensive analysis of the diffusion parameters, we identify optimal configurations maximizing adversarial robustness with minimal impact on regular performance. Importantly, this study reveals insights into the relationship between diffusion noise and diffusion steps, representing a novel contribution to the field. Our experiments are carried out on two datasets and against five adversarial attacks. The implementation code is publicly available
Bacterial antimicrobial resistance in coastal waters of two contrasting temperate environments
No full textInternational audienceBackground and objectives: The spread of antimicrobial resistance is well-known between animals and humans; however, the role of the environment, particularly in estuarine ecosystems, is less understood. Estuarine waters, downstream of catchments with human activities may be polluted by fecal bacteria carrying antibiotic-resistance genes (ARGs) and chemicals such as heavy metals, biocides or disinfectants. Due to their high filtration capacity, clams accumulate urban and agricultural bacteria, ARGs, and pollutants. Clam-related outbreaks have been reported, and the presence of antibiotic-resistant bacteria or ARGs in clams may pose additional risks to human health. This longitudinal study investigated the microbiome, resistome, and environmental factors driving the spread of ARGs in estuary to better understand their prevalence and dynamics.Methods: The presence of ARGs and integrons was investigated using high-throughput microfluidic qPCR (targeting 45 ARGs from nine antibiotic families and two integrons) in river, estuarine water, surface sediments and clams, from an anthropized site in the Bay of Brest (France), over a 14-month period (n = 258). In parallel, hydrological (river flows), physico-chemical (temperature, salinity, conductivity, nutrients...), fecal bacterial indicator (E. coli), and other pollutants (biocides, toxic metals, and antibiotics) were measured. Bacterial communities were analyzed using V3-V4 16S rRNA gene amplicon sequencing, and the resulting data were processed using SAMBA (v4.0.1). R scripts were used to explore correlations between ARGs, bacterial diversity, environmental factors and pollutant concentrations.Results 1: Fifteen ARGs conferring resistance to eight antibiotic classes, including aminoglycosides (strA, strB, ant(3')Ia), β-lactams (blaTEM, blaSHV), folic acid synthesis inhibitors (sul1, sul2, dfrA1), macrolides (mphB, mphE, ermB), phenicols (floR), polymyxins (mcr4) and tetracyclines (tetA, tetM), as well as an integron (int1), were sporadically detected in clams during the monthly monitoring. These ARGs were also found, in waters and/or sediments at higher frequencies, with other genes conferring resistance to β-lactams (blaCTX-M1, blaPER1, blaOXA23, blaOXA48, blaOXA51, blaOXA58, blaCMY1-3), fluoroquinolones (qnrA, qnrB1, qnrS), folic acid synthesis inhibitors (sul3, dfrA12, dfrA17) and macrolides (mphA). Analysis of bacterial communities showed that 24,1% of bacterial genera were shared between clams, sediments, and waters, representing 75% of the total relative abundance. Correlation analysis between the presence of ARGs and bacterial genera revealed significant associations (p < 0.01) with strong correlations (Pearson's r > 0.8). A notable example is the association of blaCTX-M1, blaSHV, mphA, and sul3 in river water, and dfrA1 and sul2 in estuarine water with Arcobacter, a ubiquitous and pathogen-related genus. This genus was detected in both river and estuarine waters, as well as in clams, with its presence increasing significantly after heavy rainfall (8.6–18.3 mm) (Pearson’s r = 0.94, p < 0.001 in river water; r = 0.88, p < 0.001 in estuarine water and r=0,48, p <0,01 in clams). In the river, its occurrence was strongly correlated with agricultural pollutants, including Metolachlor ESA (r = 0.59, p < 0.05) and Metolachlor OXA (r = 0.93, p < 0.001). In estuarine water, it was associated with both agricultural and urban pollutants, such as Metolachlor NOA and Metolachlor OXA (r = 0.75, p < 0.01), as well as E. coli concentrations (r = 0.87, p < 0.001).Conclusion: These results highlight the presence of ARGs in estuarine environments, including within clams. They also reveal strong correlations between ARGs and pathogen-related genera. These findings provide valuable insights into the potential links between environmental, human, and animal health, enhancing our understanding of the dynamics and dissemination of antimicrobial resistance in estuarine ecosystems
Représentation des risques et positionnements face aux enjeux territoriaux dans le cadre d’une comparaison internationale France-Islande
No full textInternational audienceThe aim of this paper is to study the representation of risks among coastal populations and their positioning with regard to the issues affecting their territory. These representations and positioning are studied by comparing populations of two coastal territories that differ geographically but share a strong attachment to the sea and coastline: Pays du Haut-Léon (Brittany, France) and West Fjords (Iceland). Within these two territories, we also take into account the different socio-demographic affiliations (age, gender, activity...). To meet our objective, a quantitative survey (online questionnaire) was carried out among 324 people. The results are analysed in 3 stages: the study of risk representation, the different positions taken according to the territories surveyed, and positions taken according to group status. These results show that, beyond a shared vision of local issues (ecology, housing, etc.), radically different positions on risk can emerge. These results are explained and discussed in the light of cultural differences and relations of domination between groups.Cet article a pour objectif d’étudier la représentation des risques auprès de populations littorales et leur positionnement face aux enjeux qui touchent leur territoire. Ces représentations et positionnement sont étudiés en comparant des populations de deux territoires littoraux différents du point de vue géographique mais ayant en commun un attachement fort à la mer et au littoral : le Pays du Haut-Léon (Bretagne, France) et les Fjords Ouest (Islande). Au sein de ces deux territoires, nous prenons aussi en compte les différentes appartenances socio-démographiques (âge, genre, activité, etc.). Pour répondre à notre objectif, une enquête quantitative (par questionnaire en ligne) a été menée auprès de 324 personnes. Les résultats sont analysés en 3 étapes : l’étude de la représentation des risques, les prises de position différentes selon les territoires enquêtés et les prises de position selon le statut des groupes déterminés par des caractéristiques socio-démographiques. Ces résultats mettent en évidence qu’au-delà d’une vision commune de problématiques locales (l’écologie, les logements, etc.), des positionnements radicalement différents en termes de risque peuvent se manifester. Ces résultats sont expliqués et discutés au regard des différences culturelles et de rapports de domination entre groupes
Micro-états: Sont-ils importants ?
No full textThis note addresses the relevance of rare events in system dynamics, inspired by Jill North reflections on the origin of the arrow of time in thermodynamics. After identifying the existence of rare events, characterized by a Pareto distribution, within a simple gas particle simulation, we investigate their impact on entropy evolution. These rare events are associated with microstates that locally decrease entropy, in contrast to the overall entropy increase observed in the bulk of the system. We present numerical simulations of gas particles, both without and with a gravity-like attractive force, to explore the fate of these rare events. Our results show that, while rare events can transiently generate local decreases in entropy, global entropy may continue to increase in accordance with the second law of thermodynamics. The introduction of gravity-like attraction stabilizes these low-entropy configurations, allowing them to persist longer. This study highlights the interplay between rare statistical fluctuations and macroscopic thermodynamic behavior, providing new insights into the emergence and stability of order in complex systems.Cette note aborde la pertinence d'événements rares dans la dynamique des systèmes, en s’inspirant des réflexions de Jill North sur l’origine de la flèche du temps en thermodynamique. Après avoir identifié l’existence d’événements rares, caractérisés par une distribution de Pareto, dans une simulation simple de particules de gaz, nous étudions leur impact sur l’évolution de l’entropie. Ces événements rares sont associés à des micro-états qui diminuent localement l’entropie, contrairement à l’augmentation globale de l’entropie observée dans la majeure partie du système. Nous présentons des simulations numériques de particules de gaz, sans et avec une force d’attraction de type gravitationnel, afin d’explorer le devenir de ces événements rares. Nous proposons que, bien que ces événements rares puissent générer temporairement des diminutions locales de l’entropie, l’entropie globale peut continuer à croître conformément à la seconde loi de la thermodynamique. L’introduction d’une attraction de type gravitationnel stabilise ces configurations à faible entropie, leur permettant de persister plus longtemps. Cette étude met en lumière l’interaction entre les fluctuations statistiques rares et le comportement thermodynamique macroscopique, apportant un nouvel éclairage sur l’émergence et la stabilité de l’ordre dans les systèmes complexes
Comment lire des œuvres « monuments » ? Le cas de la recherche sur Proust et Borges
No full textInternational audienceDans le cadre de mes recherches portant sur Proust et Borges, deux figures canoniques de la modernité littéraire, se pose de manière particulièrement aiguë la question de l’articulation entre distant et close reading. Cette communication propose une réflexion méthodologique sur cinq difficultés majeures rencontrées lors de l’analyse de ces œuvres dites « monuments », dont la richesse et la portée appellent à la fois une lecture exhaustive et un recul critique.La première difficulté tient à la masse même des œuvres : comment envisager une lecture « totale » – de surcroît au milieu des autres sollicitations académiques – de l’ensemble des textes proustiens et borgésiens, entre récits, essais et entretiens ? La deuxième difficulté réside dans l’arbitrage nécessaire concernant la sélection des extraits à analyser : sur quels critères choisir les passages à commenter en profondeur, en close reading ? La troisième touche à la légitimité critique : face à des bibliographies aussi vastes que spécialisées, comment faire entendre une voix propre en distant reading, notamment en littérature générale et comparée, où sont mis en dialogue des textes issus d’aires linguistiques et culturelles très différentes ?J’aborde également un problème spécifique du close reading : celui de l’analyse des traductions de Borges, parfois approximatives ou édulcorées, qui rendent difficile l’interprétation fine de certaines citations. Enfin, j’évoque les enjeux critiques liés à la « magistralité » de ces auteurs : étudier Proust et Borges, c’est aussi interroger les mécanismes de canonisation, c’est-à-dire les dynamiques culturelles, commerciales voire politiques qui tendent à « patrimonialiser » certains « faits » littéraires. Cela suppose de prendre en compte la part de subjectivité dans la lecture et dans la recherche elle-même, notamment dans le cadre de rapports intertextuels ou intersubjectifs comme celui du maître et du disciple (l’entrée thématique choisie pour ma thèse). À partir d’exemples précis, je propose donc une réflexion sur les ajustements méthodologiques rendus nécessaires par l’étude de tels corpus
Connecting the Deep Collection of Sinking Particles With Surface Ocean Signatures
No full textInternational audienceA major pathway in the biological carbon pump is the gravitational sinking of organic particles from the sunlit ocean (0–200 m) to the deep ocean. Variability in particle fluxes measured by sediment traps is often attributed to variability in primary production in the surface ocean. However, particle fluxes are also influenced by physical processes such as mesoscale eddies and fronts. In this study, we assess the impact of upper‐ocean dynamical structures on the variability of particle collection in the deep ocean. This is achieved by forward tracking the trajectories of 51.9 million virtual particles that were homogeneously released at a depth of 200 m with a constant sinking velocity of 50 m in the Northeast Atlantic basin. We found that, despite a homogeneous particle source without biological effects, purely dynamical changes can induce heterogeneity in particle density and origin at depth. The position of sediment traps can thus significantly influence the weekly to seasonal particle collection in the deep ocean. Additionally, we identify and characterize nine particle clusters using a machine‐learning approach. The results show that the seasonality of particle collection at depth can be induced by seasonal variations in upper‐ocean flow structures. Clusters associated with eddy and frontal structures are found to intermittently contribute more than 50% of the particle amount during winter and spring, with smaller secondary peaks in the summer months. This study highlights the connection between mesoscale ocean dynamics and the spatio‐temporal pattern of conservative (non‐biological) particle collection in the deep ocean
Maltodextrin-modified lipoplexes for enhanced mucosal penetration and efficient mRNA delivery
No full textInternational audienceEfficient delivery of messenger ribonucleic acid (mRNA) to mucosal tissues represents a promising approach for localized protein production in the nasal and respiratory tract. Here, we investigate the use of maltodextrin (MDX) as a surface modifier to enhance the delivery of mRNA-loaded histidylated lipoplexes (LXs) to airway epithelial cells. By reducing hydrophobicity, MDX facilitates better penetration through the mucus layer, enabling effective mRNA delivery. MDX-coated LXs improve mRNA delivery and expression in vitro by increasing cellular uptake and supporting sustained protein production. Additionally, MDX incorporation stabilizes in-house-formulated lipoplexes and modulates their interactions with mucin-covered cells. Notably, MDX-coated mRNA LXs display a four-fold increased transfection efficiency, and the protein expression is maintained up to 48 h post-transfection. Furthermore, intranasal administration of MDX-LXs results in efficient gene expression in vivo. Overall, our findings reveal that integrating MDX into mRNA lipoplexes is a promising strategy to advance nasal delivery for gene therapy and protein replacement applications
On the trail of ocean plastic
No full textPlastic owes its commercial success to its durability. But this quality becomes a flaw when it comes to plastics in the environment, where they pose toxicity problems, among others
Unravelling the proteome during sulfur compound metabolism in Dissulfuribacter thermophilus S69ᵀ, a deep-sea hydrothermal vent bacterium
No full textInternational audienceIn deep-sea hydrothermal vent environments, inorganic sulfur compounds are importantenergy substrates for sulfide/sulfur-oxidizing, sulfur/sulfate-reducing and sulfurcompounds-disproportionating microorganisms. Among these, thiosulfate disproportionatingand sulfite-reducing bacteria have been poorly studied. Moreover, themetabolic pathways and enzymatic mechanisms involved are only partially resolved.Although several genes have been proposed in the literature as possible markers for thesecatabolisms, they need to be functionally validated.Here, we tested thiosulfate disproportionation and sulfite respiration/disproportionation activity in Dissulfuribacter thermophilus S69T, an anaerobic thermophilic bacterium isolated from a deep-seahydrothermal vent at 1910 m depth at the Eastern Lau Spreading Center, Valu Fa Ridge,Pacific Ocean. We used the quantitative label free proteomics (LC-MS/MS, Tims TOFPro, Bruker) in data dependent and independent acquisition mode (DDA, DIA) to identifythe proteins involved in these catabolisms after cellular fractionation (extracellular,cytoplasmic and membrane). The tryptic peptides were obtained using SP3 samplepreparation (Hughes et al., 2019). Subsequently, peptide identification and quantificationwere performed using i2MassChroQ for DDA and DIANN for DIA. Statistical analysiswas performed using MCQR scripts developed by PAPPSO.Growth was observed under either conditions of thiosulfate disproportionation (N2/CO2) andsulfite respiration (H2/CO2)/sulfite disproprotionation. Proteomic analysis revealed multiple proteins related to the catabolisms under study. It also enabled us to functionally validate specific geneticmarkers reported in the literature and to identify new ones.These findings expand our knowledge of the sulfur catabolism in strain S69T and willimprove our understanding of the microbial contribution of this strain to the sulfur cyclein marine hydrothermal environments