Hasil untuk "Toxicology. Poisons"

Francisco Esteves, Helena Caria

The central mechanistic hypothesis underlying multifactorial hearing loss posits that genetic susceptibility and environmental exposures act synergistically to disrupt cochlear homeostasis through redox imbalance, mitochondrial dysfunction, and pro-inflammatory mechanisms. This gene–environment paradigm has significant translational implications: elucidating the molecular crosstalk between genetic variants and environmental factors may enable precision risk stratification and the development of targeted otoprotective strategies. The present review provides a comprehensive examination of the major determinants implicated in hearing loss. The manuscript is organized into six main sections that encompass the most relevant domains of current research. First, it offers (I) an overview of epidemiological patterns and the multifactorial nature of hearing impairment. This is followed by (II) a background synthesis of the complex genetic architecture underlying hearing loss. Next, the authors present (III) an outline of environmental determinants and exposure profiles associated with auditory dysfunction, highlighting prominent pollutant/xenobiotic classes (e.g., organic solvents and volatile aromatic hydrocarbons, heavy metals, pesticides, and especially organophosphates and persistent organochlorine compounds), followed by (IV) an analysis of oxidative stress, mitochondrial impairment, and inflammatory pathways involved in cochlear injury. Subsequently, (V) translational perspectives and integrated therapeutic approaches are discussed, with emphasis on epidemiological prevention and precision-based interventions. Finally, (VI) this review addresses current challenges and future directions in elucidating gene–environment interactions in hearing loss.

Samantha L Malone Rubright, L. Pearce, J. Peterson

Yuting Wu, Ziman Jiao, Yuxin Liu et al.

Abstract Background In recent years, air pollution has been increasingly recognized as a contributing factor to the onset of different diseases. This study seeks to investigate the potential link between air pollutants and dry eye (DE), as well as to identify key genes that may play a pivotal role in this relationship. Methods A comprehensive search of multiple online databases was conducted to determine target genes associated with both air pollutants and DE. Protein-protein interaction (PPI) analysis and visualization were performed for the overlapping genes. This was followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses to elucidate potential mechanisms. Subsequently, molecular docking and molecular dynamics simulation (MDS) analysis were employed to examine interactions between key genes and air pollutants. Results Eight ubiquitous air pollutants (benzene, SO₂, O₃, NO, CO, toluene, formaldehyde, and naphthalene) were chosen for analysis, and a total of 26 intersecting targets associated with DE were identified. Functional enrichment analyses using GO and KEGG pathways indicated that these targets are predominantly involved in biological processes including lipid metabolism, hormone regulation, oxidative stress, and inflammatory responses. This analysis underscored six pivotal genes linked to air pollutants and DE. Molecular docking analysis substantiated potential interactions between air pollutants and these core targets. Futhermore, MDS analysis revealed more stable relationships between MMP-2, EGFR, MMP-9, ESR-1 and air pollutants. Conclusion These findings indicate that TNF, EGFR, MMP-9, ESR-1, ICAM-1, and MMP-2 are integral to the mechanisms through which air pollutants influence DE. This research provides fresh insights about molecular mechanisms linking air pollutants and DE, emphasizing the critical need to enhance public awareness regarding the pervasive issue of air pollution.

Sahar Hamoonnavard, Hossein Rezvan, Fatemeh Behdarvand et al.

Background: The adverse effects of traditional chemical treatments have driven interest in herbal compounds for leishmaniasis therapy. Objective: This study aimed to investigate the anti-leishmanial effect of Boswellia sacra gum resin extract on the survival of Leishmania major promastigotes. Methods: Leishmania major strain (MHOM/IR/75/ER) was adjusted to 2 × 105 parasites per well. Promastigotes were exposed to a range of B. sacra gum resin extract concentrations (15, 25, 50, 70, 100, 200, 400, 800, 1600, and 3200 µg/ml) for 24 hours. Viability and cytotoxic effects were assessed using the MTT assay (3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide) with absorbance measured at 450 nm. Result: The B. sacra gum resin extract reduced promastigotes viability at various concentrations, with statistically significant effects observed at 800, 1600, and 3200 µg/mL (P ≤ 0.05). Cytotoxic effects at 1600 and 3200 µg/mL were significantly greater than those at 15, 25, and 50 µg/mL (P ≤ 0.01). Conclusion: Based on the in vitro anti-leishmanial activity observed, Boswellia sacra resin extract warrants further investigation as a potential agent against cutaneous leishmaniasis. Future in vivo studies are recommended to evaluate efficacy and safety.

Aminu Imam, Abdulraheem Mudathir Tunde, Abdulbasit Amin et al.

Abstract Background Epilepsy, characterized by recurrent seizures and associated cognitive impairments, is often exacerbated by environmental neurotoxins such as Cypermethrin. This study investigates the effects of Cypermethrin (CPM) on Pentylenetetrazole (PTZ)-induced seizures and the potential neuroprotective properties of Vitamin E (VIT E) and valproate (VAP), both alone and in combination. Methods Behavioral tests, biochemical assays, and immunohistochemical analyses were conducted to assess seizure severity, cognitive performance, neuronal integrity, and markers of neuroinflammation and oxidative stress. Results PTZ-induced seizures significantly increased seizure severity and impaired cognitive performance in the T-maze, Y-maze, and Morris’ water maze tests, with these deficits being exacerbated by CPM. Immunohistochemical staining revealed substantial neuronal and interneuron loss in the hippocampus, accompanied by heightened microglial activation. Biochemical analyses showed elevated levels of pro-inflammatory cytokines (TNF-α) and oxidative stress markers (MDA), alongside reduced antioxidant enzyme activity (SOD). Therapeutic intervention with VAP or VIT E mitigated these effects to varying degrees, demonstrating improvements in seizure scores, cognitive performance, neuronal integrity, and reductions in neuroinflammation and oxidative stress markers. These findings underscore the potential synergistic benefits of combining VAP and VIT E in managing seizures and associated neuropathologies. Conclusion This study highlights the exacerbating effects of environmental toxins on seizure-related cognitive and neuronal impairments while providing evidence for antioxidant-based combination therapies as a promising approach to mitigate these effects.

Meng LIU, Wei ZHANG, Qi ZHUANG et al.

BackgroundThe high work intensity and possible subsequently increased susceptibility to occupational hazards of calcium carbide plants may lead to hypertension in workers, but there are few studies on the relationship between occupational hazard exposure and hypertension in workers involving the production process of calcium carbide.ObjectiveTo explore the influence of occupational hazards on the incidence of hypertension in calcium carbide plants.MethodsUsing historical cohort design, the employees of a calcium carbide factory in the western part of Inner Mongolia Autonomous Region were selected as research subjects. According to the pre-determined inclusion and exclusion criteria, the study population comprised an exposure group of 377 employees (including furnace workers, inspection workers, and maintenance workers) exposed to dust, noise & carbon monoxide, and a control group of 388 employees (including central control workers, electricians, and administrative personnel) without above-mentioned exposure. The total sample size was 765 participants. The follow-up period was from April 2011 to October 2022, and the study endpoint was defined as the conclusion of the follow-up period or diagnosed hypertension in annual occupational health examination. Information on general demographic characteristics, living habits, and work status was collected from all study subjects. Cox proportional hazards regression model was used to analyze the association between occupational hazard exposure and the risk of hypertension among the calcium carbide plant employees.ResultsThe average age, mean systolic and diastolic blood pressure, proportion of males, smoking rate, and alcohol consumption rate in the exposure group were higher than those in the control group (P<0.05). Compared to baseline, both systolic and diastolic blood pressure levels increased in the exposure group and the control group at the end of the follow-up (P<0.05). At the end of the follow-up, the average differences between systolic/ diastolic blood pressure and baseline values in the exposure group were higher than those in the control group (P<0.05). During the follow-up period, a total of 223 cases of hypertension occurred, with a total follow-up of 2785 person-years, resulting in an incidence density of 8007.18 per 100000 person-years, an average age of onset of (35.90±8.22) years, and an average working age of onset of (2.69±1.97) years. The incidence density in the exposure group was 128.71% higher than that in the control group, and the risk of hypertension in the exposure group was 2.115 times that of the control group. The risk of hypertension was 2.199 times higher for men than for women, 1.344 times higher for those aged 30 and above than for those under 30, 1.546 times higher for smokers than for non-smokers, and 1.750 times higher for drinking workers than for non-drinking ones. The results of Cox proportional hazards regression modeling indicated that the hazard ratio (95%CI) of hypertension among the employees exposed to dust, noise, and carbon monoxide was 2.254 (1.703, 2.982), 1.594 (1.107, 2.295), and 1.567 (1.079, 2.274), respectively, when different covariates (gender, age, smoking, and alcohol consumption) were included. The results of Log-rank test showed that there were statistically significant differences in the distribution of disease-free survival between the exposure group and the control group (P<0.05).ConclusionOccupational exposures to dust, noise, and carbon monoxide may increase the risk of hypertension among calcium carbide plant workers.

William Xu, Yiwei Lu, Yihan Wang et al.

Targeted data poisoning attacks pose an increasingly serious threat due to their ease of deployment and high success rates. These attacks aim to manipulate the prediction for a single test sample in classification models. Unlike indiscriminate attacks that aim to decrease overall test performance, targeted attacks present a unique threat to individual test instances. This threat model raises a fundamental question: what factors make certain test samples more susceptible to successful poisoning than others? We investigate how attack difficulty varies across different test instances and identify key characteristics that influence vulnerability. This paper introduces three predictive criteria for targeted data poisoning difficulty: ergodic prediction accuracy (analyzed through clean training dynamics), poison distance, and poison budget. Our experimental results demonstrate that these metrics effectively predict the varying difficulty of real-world targeted poisoning attacks across diverse scenarios, offering practitioners valuable insights for vulnerability assessment and understanding data poisoning attacks.

Nilufer Gulciftci, M. Emre Gursoy

Poisoning attacks, in which an attacker adversarially manipulates the training dataset of a machine learning (ML) model, pose a significant threat to ML security. Beta Poisoning is a recently proposed poisoning attack that disrupts model accuracy by making the training dataset linearly nonseparable. In this paper, we propose four defense strategies against Beta Poisoning attacks: kNN Proximity-Based Defense (KPB), Neighborhood Class Comparison (NCC), Clustering-Based Defense (CBD), and Mean Distance Threshold (MDT). The defenses are based on our observations regarding the characteristics of poisoning samples generated by Beta Poisoning, e.g., poisoning samples have close proximity to one another, and they are centered near the mean of the target class. Experimental evaluations using MNIST and CIFAR-10 datasets demonstrate that KPB and MDT can achieve perfect accuracy and F1 scores, while CBD and NCC also provide strong defensive capabilities. Furthermore, by analyzing performance across varying parameters, we offer practical insights regarding defenses' behaviors under varying conditions.

R. Slaughter, M. Watts, J. Allister Vale et al.

Abstract Introduction: Sodium hypochlorite is used as a bleaching and disinfecting agent and is commonly found in household bleach. Objective: The objective is to review critically the epidemiology, mechanisms of toxicity, clinical features, diagnosis, and management of hypochlorite poisoning. Methods: PubMed was searched from January 1950 to June 2018 using the terms “Hypochlorite”, “Sodium Hypochlorite”, “Sodium Oxychloride”, “Hypochlorous Acid”, “Bleach”, “Chlorine Bleach”, in combination with the keywords “poisoning”, “poison”, “toxicity”, “ingestion”, “adverse effects”, “overdose”, and “intoxication”. In addition, bibliographies of identified articles were screened for additional relevant studies including non-indexed reports. Non-peer-reviewed sources were also included. These searches produced 110 citations which were considered relevant. Epidemiology: There is limited information regarding statistical trends on world-wide poisoning from sodium hypochlorite. In the United States of America, poison control center data have shown that enquiries regarding hypochlorite bleaches have ranged from 43,000 to 46,000 per year over the period 2012–2016. Mechanisms of toxicity: Hypochlorite’s potential to cause toxicity is related to its oxidizing capacity and the pH of the solution. Toxicity arises from its corrosive activity upon contact with mucous membranes and skin. Features following ingestion: While small accidental ingestions are very unlikely to cause clinically significant toxicity, large ingestions may cause corrosive gastrointestinal injury and systemic effects, including metabolic acidosis, hypernatremia, and hyperchloremia. Features following dental exposure: Hypochlorite is used extensively by dentists for cleaning root canals and is safe if the solution remains within the root canal. Extrusions into the periapical area can result in severe pain with localized large and diffuse swelling and hemorrhage. Features following skin exposure: Prolonged or extensive exposure may cause skin irritation and damage to the skin or dermal hypersensitivity. Such exposures can result in either immediate or delayed-type skin reactions. High concentration solutions have caused severe chemical skin burns. Features following inhalation: Although there are only limited data, inhalation of hypochlorite alone is likely to lead to no more than mild irritation of the upper airways. Features following ocular exposure: Corneal injuries from ocular exposure are generally mild with burning discomfort and superficial disturbance of the corneal epithelium with recovery within 1 or 2 days. With higher concentration solutions, severe eye irritation can occur. Diagnosis: The diagnosis can typically be made on the basis of a careful history, including details of the specific product used, its hypochlorite concentration, and the amount involved. As hypochlorite bleach produces a characteristic smell of chlorine, this may provide a diagnostic clue. In severe cases, corrosive injury is suggested on presentation because of hypersalivation, difficulty swallowing, retrosternal pain or hematemesis. Management: Symptom-directed supportive care is the mainstay of management. Gastrointestinal decontamination is not beneficial. Local corrosive injury is the major focus of treatment in severe cases. Fiberoptic endoscopy and CT thorax/abdomen are complimentary and have been shown to be useful in corrosive injuries in assessing the severity of injury, risk of mortality and risk of subsequent stricture formation and should be performed as soon as possible after ingestion. Dental periapical extrusion injuries should be left open for some minutes to allow bleeding through the tooth and to limit hematoma development in tissue spaces. Once the bleeding has ceased, the canal can be dressed with non-setting calcium hydroxide and sealed coronally. Conclusions: Accidental ingestion of household bleach is not normally of clinical significance. However, those who ingest a large amount of a dilute formulation or a high concentration preparation can develop severe, and rarely fatal, corrosive injury so prompt supportive care is essential as there is no specific antidote. Treatment primarily consists of symptom-directed supportive care.

Richard C. Semelka, Miguel Ramalho

PurposeTo demonstrate and evaluate factors contributing to near-cures in patients with Gadolinium Deposition Disease (GDD) undergoing intravenous (IV) DTPA chelation.MethodsPatients who had undergone or are currently undergoing DTPA chelation for GDD were included in this report based on their medical records that showed their perceived improvement was at least 80% back to normal. A survey was developed that included factors commonly reported by patients treated in one clinic to determine if these ‘near-cured’ (pre-MRI baseline health) individuals possessed certain factors and lacked others. The anonymized survey was emailed to these individuals by the principal treating physician, the only investigator not blinded to the subjects. This report describes clinical documentation of patient status and their underlying factors in individuals treated by the primary author, and no research was performed. The survey was sent to sixteen individuals; Fourteen patients completed it (10 females; 41.1 ± 11.2 y/o).ResultsThe most common factor was the administration of ≤5 lifetime doses of a Gadolinium-Based Contrast Agents (GBCA) (12/14). Unconfounded agents triggering GDD were seen in nine subjects. Most subjects (12/14) initiated chelation in the first year after the causative GBCA, and most (11/14) underwent ≤10 chelations with DTPA. Good healthcare status prior to MRI was observed in 5 subjects. The majority (11/14) described their immune status as strong. Severe physical disability prior to chelation was seen in 1.ConclusionSubjects with GDD can experience near-cure with IV DTPA chelation. Factors surveyed that predict near-cure include the start of chelation in the first year, few GBCA administrations, and good health status before MRI with GBCA injection. Nonetheless, a few patients with predictors of less successful outcomes still experienced near-cure.

Di WANG, Xiaoni CHI, Zishan HUANG et al.

Climate change has led to an increasing frequency and intensity of extreme climate events such as heat and drought extremes with considerable global public health burden. This systematic review collected 87 domestic and international studies from 2000 to 2023, considering the impacts of heat extremes, drought extremes, and compound hot-dry events on infectious diseases attributable to various transmission pathways such as waterborne, foodborne, insect-borne, airborne, and contact-transmitted diseases. Our results showed that high temperature was associated with increased transmission risks of waterborne and foodborne diseases including infectious diarrheal diseases (cholera, dysentery, typhoid, and paratyphoid) and infectious gastroenteritis; vector-borne diseases including dengue fever, Zika virus (ZIKV) disease, chikungunya fever, malaria, West Nile fever, and Rift Valley fever; airborne diseases including influenza-like diseases, influenza A, measles, and mumps; and contact-transmitted diseases including HIV/AIDS, schistosomiasis, and leptospirosis. Additionally, drought conditions also amplified the transmission risks of waterborne and foodborne diseases including cholera, Escherichia coli infection, rotavirus infection, and hepatitis E; vector-borne diseases such as scrub typhus, schistosomiasis, hemorrhagic fever with renal syndrome, and West Nile fever; airborne diseases including meningococcal meningitis, pertussis, measles, and upper respiratory infections; and contact-transmitted diseases such as HIV/AIDS. Along with global warming, the frequency of compound high temperature and drought events shows a considerably increasing trend, causing more adverse health effects than heat or drought alone. However, there is limited research quantifying their effects on infectious diseases. These associations may be mediated through temperature and precipitation on infectious disease pathogens, transmission vectors, population susceptibility, public health services, and behaviors. In the context of climate change, the increasing occurrence of compound events of high temperatures and droughts raises health concerns, and further studies are needed to enhance our understanding of the impacts of climate change on infectious diseases and improve human adaption to climate change.

Yuqi Jia, Minghong Fang, Hongbin Liu et al.

Poisoning attacks compromise the training phase of federated learning (FL) such that the learned global model misclassifies attacker-chosen inputs called target inputs. Existing defenses mainly focus on protecting the training phase of FL such that the learnt global model is poison free. However, these defenses often achieve limited effectiveness when the clients' local training data is highly non-iid or the number of malicious clients is large, as confirmed in our experiments. In this work, we propose FLForensics, the first poison-forensics method for FL. FLForensics complements existing training-phase defenses. In particular, when training-phase defenses fail and a poisoned global model is deployed, FLForensics aims to trace back the malicious clients that performed the poisoning attack after a misclassified target input is identified. We theoretically show that FLForensics can accurately distinguish between benign and malicious clients under a formal definition of poisoning attack. Moreover, we empirically show the effectiveness of FLForensics at tracing back both existing and adaptive poisoning attacks on five benchmark datasets.

Tingchen Fu, Mrinank Sharma, Philip Torr et al.

Preference learning is a central component for aligning current LLMs, but this process can be vulnerable to data poisoning attacks. To address this concern, we introduce PoisonBench, a benchmark for evaluating large language models' susceptibility to data poisoning during preference learning. Data poisoning attacks can manipulate large language model responses to include hidden malicious content or biases, potentially causing the model to generate harmful or unintended outputs while appearing to function normally. We deploy two distinct attack types across eight realistic scenarios, assessing 21 widely-used models. Our findings reveal concerning trends: (1) Scaling up parameter size does not inherently enhance resilience against poisoning attacks; (2) There exists a log-linear relationship between the effects of the attack and the data poison ratio; (3) The effect of data poisoning can generalize to extrapolated triggers that are not included in the poisoned data. These results expose weaknesses in current preference learning techniques, highlighting the urgent need for more robust defenses against malicious models and data manipulation.

Z. Svirčev, Damjana Drobac, Nada Tokodi et al.

Jing WU, Cuiyun ZUO, Yanyan KE et al.

BackgroundOsteoclast stimulatory transmembrane protein (OC-STAMP) is involved in silicosis fibrosis induced by silicon oxide (SiO2) exposure. Its role in silicosis fibrosis by inducing ferroptosis of alveolar type II epithelial cells and its related mechanism remain unclear. ObjectiveTo explore the effect and possible mechanism of OC-STAMP on ferroptosis of alveolar type II epithelial cells and silicosis fibrosis in rats under SiO2 exposure. MethodsTwenty male Wistar rats of SPF grade were randomly divided into two groups: control (Sham) group and SiO2 group, 15 rats in each group. Rats in the SiO2 group were given 1 mL of 50 mg·L−1 SiO2 suspension at one time through the non-exposed intratracheal instillation method to establish an animal model of silicosis, and rats in the Sham group were give 1 mL of 0.9% sodium chloride solution in the same way. Rats were sacrificed after 8 weeks. Samples of lung tissue were fixed in glutaraldehyde or paraformaldehyde for observing ultrastructure of mitochondria by transmission electron microscopy; HE, Masson, VG, and Prussian blue were used to observe changes in lung tissue structure and iron deposition. The expression level of OC-STAMP and the degree of lung fibrosis were evaluated by immunohistochemistry and immunofluorescence. The expression level of OC-STAMP in rat lung tissue was detected and the transfection effect of OC-STAMP was verified by real-time fluorescence quantitative polymerase chain reaction (RT-PCR). Overexpression (OCS group) and inhibition expression (SI-OC group) models were constructed by OC-STAMP plasmid and OC-STAMP small interfering RNA (siRNA) transfection to cultured MLE-12 cells, respectively. The relative expression levels of glutathione peroxidase 4 (GPX4), solute carrier family 7 member 11 (SLC7A11), and other proteins in lung tissue and MLE-12 were detected by Western blotting. ResultsThe results of HE, Masson, and VG staining showed that the silicosis modeling was successful after 8 weeks of SiO2 exposure. The immunofluorescence results showed that OC-STAMP and ATP binding cassette subfamily A member 3 (ABCA3) co-localized in alveolar type II epithelium. The immunohistochemical results showed that the levels of OC-STAMP and collagen I in the SiO2 group were significantly higher than those in the Sham group (P<0.01). The RT-PCR results showed that the OC-STAMP mRNA in the lung tissue of the SiO2 group was significantly higher than that of the Sham group (P<0.01). The Prussian blue staining in the lung tissue of the SiO2 group showed positive brownish-yellow particles. Compared with the Sham group which showed normal mitochondrial structure, the mitochondrial structure was generally swollen and the mitochondrial cristae dissolved and disappeared in the SiO2 group by transmission electron microscope observation. The Western blotting results showed that the expression levels of SLC7A11 and GPX4 both decreased in the lung tissue of the SiO2 group (P<0.05, P<0.01), and the expression level of Vimentin increased (P<0.01). In the transfected MLE-12 cells, compared with the Sham group, the expression levels of SLC7A11 and GPX4 in the OCS group were significantly reduced (P<0.05, P<0.01). ConclusionOC-STAMP may affect the expression of proteins related to ferroptosis, and promote lung fibrosis induced by SiO2 exposure.

Jingyi HE, Yan ZHOU, Xiao LIU et al.

BackgroundPerfluoroalkyl and polyfluoroalkyl substances (PFASs) exposure may affect male reproductive health. There are regional differences in PFASs exposure levels among men of childbearing age in China, and current research results on associated influencing factors are inconsistent. ObjectiveTo investigate the levels of PFASs in serum and their determinants among men of childbearing age. MethodsThe participants (n=113, 22-45 years old) were from a cross-sectional study of exposure to environmental pollutants and male reproductive health damage in Hubei Province conducted from April to June 2013 at the Reproductive Medicine Center of Tongji Hospital, Wuhan, Hubei Province. Eleven kinds of PFASs were measured in serum by isotopic dilution-high performance liquid chromatography-tandem mass spectrometry. The included PFASs were prefluorohexanoic acid (PFHxA), perfluoroheptanoic acid (PFHpA), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroundecanoic acid (PFUdA), perfluorododecanoic acid (PFDoA), perfluorotridecanoic acid (PFTrDA), perfluorotetradecanoic acid (PFTeDA), perfluorohexane sulfonate (PFHxS), and perfluorooctane sulfonate (PFOS). Information about participants' demographic characteristics, lifestyle, and habits was collected by a set of self-designed questionnaires. The associations of demographic characteristics, lifestyle, and habits with exposure to PFASs were analyzed by linear regression. ResultsThe major components of PFASs were PFOS and PFOA, and the concentrations expressed as M (P25, P75) were 8.31 (4.90, 17.79) ng·mL−1 and 2.77 (2.18, 3.46) ng·mL−1, respectively. The positive rates of six PFASs (PFOA, PFNA, PFDA, PFUdA, PFHxS, and PFOS) were 100%, followed by PFDoA and PFTrDA (87.61% and 88.59%, respectively). The linear regression results showed that age was positively associated with the levels of Σ8PFASs (sum of the concentrations of the eight PFASs with a positive rate greater than 80%) (P < 0.05). The concentration of serum PFOA in men with monthly family income of 2000-4000 yuan was 53.73% (P < 0.01) higher than those in men with monthly family income of <2000 yuan. The serum concentrations of PFNA and PFTrDA were both 32.31% (P < 0.05) higher in men with monthly family income ≥4000 yuan than those in men with monthly family income <2000 yuan. The serum concentration of PFHxS in men who used plastic cups was 33.64% (P < 0.01) higher than that in men who did not report oral contact with plastic products (plastic tableware, plastic cups, and plastic bags for packing food). The serum concentrations of PFHxS, PFOS, and Σ8PFASs were 33.64% (P < 0.01), 43.33% (P < 0.01), and 36.34% (P < 0.05) higher in men who bathed with laundry soap than those in men who did not use detergents. Men who bathed with toilet soap had a 34.99% (P < 0.05) higher serum concentration of PFHxS than those who bathed without detergents. ConclusionMen of childbearing age are exposed to PFASs extensively. Age, monthly household income, usage of laundry soap or toilet soap in bathing, and usage of plastic cups may influence the level of PFASs in serum. However, further investigation is needed to confirm these results.

Zijian Wang, Shuo Huang, Yujin Huang et al.

In recent years, on-device deep learning has gained attention as a means of developing affordable deep learning applications for mobile devices. However, on-device models are constrained by limited energy and computation resources. In the mean time, a poisoning attack known as sponge poisoning has been developed.This attack involves feeding the model with poisoned examples to increase the energy consumption during inference. As previous work is focusing on server hardware accelerators, in this work, we extend the sponge poisoning attack to an on-device scenario to evaluate the vulnerability of mobile device processors. We present an on-device sponge poisoning attack pipeline to simulate the streaming and consistent inference scenario to bridge the knowledge gap in the on-device setting. Our exclusive experimental analysis with processors and on-device networks shows that sponge poisoning attacks can effectively pollute the modern processor with its built-in accelerator. We analyze the impact of different factors in the sponge poisoning algorithm and highlight the need for improved defense mechanisms to prevent such attacks on on-device deep learning applications.

Yiwei Lu, Gautam Kamath, Yaoliang Yu

Indiscriminate data poisoning attacks aim to decrease a model's test accuracy by injecting a small amount of corrupted training data. Despite significant interest, existing attacks remain relatively ineffective against modern machine learning (ML) architectures. In this work, we introduce the notion of model poisoning reachability as a technical tool to explore the intrinsic limits of data poisoning attacks towards target parameters (i.e., model-targeted attacks). We derive an easily computable threshold to establish and quantify a surprising phase transition phenomenon among popular ML models: data poisoning attacks can achieve certain target parameters only when the poisoning ratio exceeds our threshold. Building on existing parameter corruption attacks and refining the Gradient Canceling attack, we perform extensive experiments to confirm our theoretical findings, test the predictability of our transition threshold, and significantly improve existing indiscriminate data poisoning baselines over a range of datasets and models. Our work highlights the critical role played by the poisoning ratio, and sheds new insights on existing empirical results, attacks and mitigation strategies in data poisoning.

Young Wu, Jeremy McMahan, Xiaojin Zhu et al.

We characterize offline data poisoning attacks on Multi-Agent Reinforcement Learning (MARL), where an attacker may change a data set in an attempt to install a (potentially fictitious) unique Markov-perfect Nash equilibrium for a two-player zero-sum Markov game. We propose the unique Nash set, namely the set of games, specified by their Q functions, with a specific joint policy being the unique Nash equilibrium. The unique Nash set is central to poisoning attacks because the attack is successful if and only if data poisoning pushes all plausible games inside the set. The unique Nash set generalizes the reward polytope commonly used in inverse reinforcement learning to MARL. For zero-sum Markov games, both the inverse Nash set and the set of plausible games induced by data are polytopes in the Q function space. We exhibit a linear program to efficiently compute the optimal poisoning attack. Our work sheds light on the structure of data poisoning attacks on offline MARL, a necessary step before one can design more robust MARL algorithms.

Sanghak Oh, Kiho Lee, Seonhye Park et al.

AI-powered coding assistant tools have revolutionized the software engineering ecosystem. However, prior work has demonstrated that these tools are vulnerable to poisoning attacks. In a poisoning attack, an attacker intentionally injects maliciously crafted insecure code snippets into training datasets to manipulate these tools. The poisoned tools can suggest insecure code to developers, resulting in vulnerabilities in their products that attackers can exploit. However, it is still little understood whether such poisoning attacks against the tools would be practical in real-world settings and how developers address the poisoning attacks during software development. To understand the real-world impact of poisoning attacks on developers who rely on AI-powered coding assistants, we conducted two user studies: an online survey and an in-lab study. The online survey involved 238 participants, including software developers and computer science students. The survey results revealed widespread adoption of these tools among participants, primarily to enhance coding speed, eliminate repetition, and gain boilerplate code. However, the survey also found that developers may misplace trust in these tools because they overlooked the risk of poisoning attacks. The in-lab study was conducted with 30 professional developers. The developers were asked to complete three programming tasks with a representative type of AI-powered coding assistant tool, running on Visual Studio Code. The in-lab study results showed that developers using a poisoned ChatGPT-like tool were more prone to including insecure code than those using an IntelliCode-like tool or no tool. This demonstrates the strong influence of these tools on the security of generated code. Our study results highlight the need for education and improved coding practices to address new security issues introduced by AI-powered coding assistant tools.