Data poisoning attacks on clustering algorithms have received limited attention, with existing methods struggling to scale efficiently as dataset sizes and feature counts increase. These attacks typically require re-clustering the entire dataset multiple times to generate predictions and assess the attacker's objectives, significantly hindering their scalability. This paper addresses these limitations by proposing Sonic, a novel genetic data poisoning attack that leverages incremental and scalable clustering algorithms, e.g., FISHDBC, as surrogates to accelerate poisoning attacks against graph-based and density-based clustering methods, such as HDBSCAN. We empirically demonstrate the effectiveness and efficiency of Sonic in poisoning the target clustering algorithms. We then conduct a comprehensive analysis of the factors affecting the scalability and transferability of poisoning attacks against clustering algorithms, and we conclude by examining the robustness of hyperparameters in our attack strategy Sonic.
ABSTRACT Two types of white wheat bread (high- and low-type loaves) were investigated for rope spoilage. Thirty of the 56 breads tested developed rope spoilage within 5 days; the high-type loaves were affected by rope spoilage more than the low-type loaves. Sixty-one Bacillus strains were isolated from ropy breads and were characterized on the basis of their phenotypic and genotypic traits. All of the isolates were identified as Bacillus subtilis by biochemical tests, but molecular assays (randomly amplified polymorphic DNA PCR assay, denaturing gradient gel electrophoresis analysis, and sequencing of the V3 region of 16S ribosomal DNA) revealed greater Bacillus species variety in ropy breads. In fact, besides strains of B. subtilis, Bacillus licheniformis , Bacillus cereus , and isolates of Bacillus clausii and Bacillus firmus were also identified. All of the ropy Bacillus isolates exhibited amylase activity, whereas only 32.4% of these isolates were able to produce ropiness in bread slices after treatment at 96°C for 10 min. Strains of lactic acid bacteria previously isolated from sourdough were first selected for antirope activity on bread slices and then used as starters for bread-making experiments. Prevention of growth of approximately 10 4 rope-producing B. subtilis G1 spores per cm 2 on bread slices for more than 15 days was observed when heat-treated cultures of Lactobacillus plantarum E5 and Leuconostoc mesenteroides A27 were added. Growth of B. subtilis G1 occurred after 7 days in breads started with Saccharomyces cerevisiae T22, L. plantarum E5, and L. mesenteroides A27.
One hundred and twenty-five isolates of Micrococcaceae from Italian salami were tested for antagonistic activities against Listeria monocytogenes. Four isolates, identified as Staphylococcus xylosus, inhibited the growth of all five strains of L. monocytogenes tested. The antagonistic substances produced by strains 39A and 41A were inactivated by some proteases, whereas those from strains 1E and 27E were inactivated only by esterase and lipase. They are neither bacteriophages, nor lytic enzymes like lysostaphin.
One hundred three isolates of Carnobacterium spp. from raw meat were analyzed by random amplification of polymorphic DNA (RAPD) and PCR and were identified by 16S rRNA gene sequencing. Forty-five strains of Carnobacterium maltaromaticum were characterized for their growth capabilities at different temperatures, NaCl concentrations, and pH values and for in vitro lipolytic and proteolytic activities. Moreover, their spoilage potential in meat was investigated by analyzing the release of volatile organic compounds (VOCs) in meat stored in air or vacuum packs. Almost all the strains were able to grow at 4, 10, and 20°C, at pH values of 6 to 9, and in the presence of 2.5% NaCl. The release of VOCs by each strain in beef stored at 4°C in air and vacuum packs was evaluated by headspace solid-phase microextraction (HS-SPME)-gas chromatography-mass spectrometry (GC-MS) analysis. All the meat samples inoculated and stored in air showed higher numbers of VOCs than the vacuum-packed meat samples. Acetoin, 1-octen-3-ol, and butanoic acid were the compounds most frequently found under both storage conditions. The contaminated meat samples were evaluated by a sensory panel; the results indicated that for all sensory odors, no effect of strain was significant (P > 0.05). The storage conditions significantly affected (P < 0.05) the perception of dairy, spoiled-meat, and mozzarella cheese odors, which were more intense in meat stored in air than in vacuum packs but were never very intense. In conclusion, different strains of C. maltaromaticum can grow efficiently in meat stored at low temperatures both in air and in vacuum packs, producing volatile molecules with low sensory impacts, with a negligible contribution to meat spoilage overall.
