Mendeley Data Showcase

Dataset Composition Examples: 4,090,661 Legitimate: 998,313 DGA: 3,092,348 DGA Families: 160 Morphological Types: 5 Dataset composed of a collection from the DGA DGArchive feed between 09/28/2025 and 10/28/2025, plus the Majestic Million domain set collected on 09/28/2025. Examples found in both lists were disregarded in order to improve the quality and reliability of the labels presented by the dataset. This dataset was conceived during research work for the detection and classification of DGA using deep learning and natural language processing techniques. This research resulted in the publication of two articles and a master's thesis: - Class Incremental Deep Learning: A Computational Scheme to Avoid Catastrophic Forgetting in Domain Generation Algorithm Multiclass Classification. https://doi.org/10.3390/app14167244 - Deep Convolutional Neural Network and Character Level Embedding for DGA Detection. http://dx.doi.org/10.5220/0012605700003690 - Detecção de domínios gerados por algoritmos com aprendizado profundo incremental e DNS passivo. https://hdl.handle.net/11449/313556 The DGA example set, we organized it by families. This data already came from the original set obtained from the DGArchive. In type, we identified five major morphological groups, which are described below: - Random: DGA families that generate their domain names in a way that forms unintelligible character sequences. Although we know that this sequence is not exactly random, it gives the impression of being random, hence the choice of the label for the morphological type. Examples: 15zkgsh1n100ax15m265x1cnkdk7.org, uhovosxkjkcg.ru - Pseudo-words: This morphological format seeks to emulate the formation of real words, permuting vowels and consonants. This technique seeks to evade security layers that are based on entropy to detect DGA domains. Examples: stajaq.com, nonafudazage.name - Seed-domain: Some DGAs start with an existing, often legitimate, domain name and add characters at the beginning or end, allow characters, and change the initial domain's TLD. Examples: tlzeitudeconscientiousavl.com, agirtvolveras.com - Wordlist: Another very effective way that some threats have found to evade automated DGA detection analyses is by forming their domains using dictionaries of real words and creating domains with the permission of those words. These DGAs are particularly difficult to detect because of their morphology, which is very close to legitimate domains. Examples: kneemanualgirl.com, leadunabledeal.art, christinashaquila.ru - Subdomain: Another strategy recently used by attackers to try to evade automated DGA analysis is the use of a subdomain instead of the initial domain. Some even use subdomains of legitimate Dynamic DNS providers. We found DGA families that have examples of both morphologies (domain and subdomain), but there are already threats that rely exclusively on this morphology. Examples: odtzcjajsrxh.dyndns.org, eboxmj56grwjs2afs6i.ddns.net

This dataset is associated with the clinical research manuscript "Chidamide: A new exploration of maintenance therapy for DLBCL patients with HBV infection." The dataset includes baseline data, pathological information, efficacy evaluation, long-term follow-up results for all DLBCL patients enrolled in the study, as well as hematologic and non-hematologic adverse events for HBV+ DLBCL patients receiving Chidamide maintenance therapy. The spreadsheet should be read from left to right, with each row recording the treatment course of each patient. The first row describes the content of each column. For any questions regarding the dataset, please contact the corresponding author of this manuscript.

Original TEM data for“Million-Year Scale Meteoroid Flux Homogeneity Across the Moon: Evidence from Chang'E-6 Samples and Gardening Simulations”

"1.xlsx" contains the main model data, including the stock price returns of Bimbo, Yonghui, and Zhongbai. "mid_down ADF" is the data with additional external control variables added for robustness testing. "CARCH_Copula_CoVaR" is the main model code, which is used to quantify the dynamic credit risk spillover rate of the supply chain; the rest are codes for robustness testing.

This study investigated the effects of partial sleep restriction on the recovery process following exercise-induced muscle damage (EIMD) in healthy, untrained men. Using a randomized crossover design, sixteen participants underwent two experimental conditions: habitual sleep (HS) and sleep restriction (SR), the latter involving a 50% reduction in total sleep time for two consecutive nights. Muscle damage was induced through an eccentric exercise protocol consisting of elbow flexion with high overload (130% of one-repetition maximum). Various markers of muscle recovery were evaluated at baseline, immediately post-exercise, and 24 and 48 hours after the protocol. These included biochemical markers (creatine kinase, lactate dehydrogenase), morphological assessments (muscle thickness, echo intensity), bioelectrical impedance (phase angle), and perceptual measures (pain sensitivity, pain tolerance, and strength performance). The results showed significant time effects for all variables, confirming that the exercise protocol effectively induced muscle damage. However, no significant differences were observed between the sleep restriction and habitual sleep conditions, indicating that two nights of 50% sleep reduction did not impair the peripheral recovery process. These findings suggest that short-term partial sleep loss may not substantially disrupt muscle repair or inflammatory modulation, although longer or repeated periods of sleep restriction could lead to cumulative physiological consequences. The study contributes novel evidence to the understanding of the interaction between sleep and skeletal muscle recovery, emphasizing that mild sleep restriction—common in daily life—may not compromise short-term muscular recovery after eccentric exercise.

This dataset originates from a pilot study examining Chilean teacher educators’ self-perceived digital competencies using an adapted DigComp 2.2 framework. The research hypothesis proposed that Digital Content Creation and Digital Security would be the weakest competency areas, based on international evidence. The dataset includes responses from 62 teacher educators collected via an online questionnaire comprising 21 items distributed across five domains: Information and Digital Literacy, Communication and Collaboration, Digital Content Creation, Security, and Problem Solving. Each item used an 8-point progressive scale corresponding to DigComp proficiency levels. The data reveal a pronounced asymmetry between digital consumption and production skills: while participants scored highly in Information and Digital Literacy (82.26% medium–high levels), performance in Security (37.10%), Content Creation (43.55%), and Problem Solving (41.94%) was significantly lower (p < .001, r = .78–.80). Cronbach’s α = .969 indicates excellent internal consistency of the instrument. Chi-square analyses (V = .45–.78) show strong interdependence among the weakest domains, suggesting that deficits cluster rather than appear independently. The dataset therefore provides early evidence of an interconnected deficit profile where low performance in one critical area predicts weakness in others. These findings can be interpreted as indicators of limited technological autonomy and creative digital agency among teacher educators. The data are valuable for researchers and policymakers aiming to design integrated professional development programs that address multiple digital competencies simultaneously. All data were gathered anonymously with informed consent. They can be reused for replication studies, cross-cultural comparisons, or instrument validation in other higher education contexts.

Dataset for preparing the artice "Demographic Aging and Government Debt: A Panel Data Analysis of 61 Countries"

All NDVI data and statistics for vegetation in Santa Lucía and Valle de Ángeles from 1985 to 2025, plus 4 residents interviewed who agreed to share their recorded voices (Spanish and English) expressing their views on excessive urbanization in the rural towns.

This repository contains raw data and analysis files relating to Roux, Lindell et al (2025) Nature Microbiology.

This repository provides the replication package for the paper “Assessing the Meta-Laws of Scientific Existence” (v1.0.0, 2025-11-07). ZIP filename: meta-laws-repro-v1_20251107.zip SHA-256: f3d5b8fde5734b26ecb09cc39e0b75e208a8dcd68146a4ff0f5512596f54c664 Contents - code/: Python 3 scripts (stability_check.py, make_diff.py, filter_1990.py) - data/: timeseries_diff_1990.csv (monthly Δlog×100; 1990–present) - protocols/: Param-Obs-Protocol.csv (ASCII, operational thresholds) - out_1990/: expected outputs (summary.md, rolling_cdf_ks.png) - checksums/: CHECKSUMS.txt (hashes), env.txt (Python version and pip freeze) - README.md: quick start and troubleshooting How to reproduce (Windows, from the extracted root) py -3 code\stability_check.py --in data\timeseries_diff_1990.csv --out out_1990 --acf-lags 36 --ks-window 120 --ks-ref 120 --ks-step 6 Expected results (our run, N=429) ADF = -4.044 (p = 0.0012) rejects a unit root; KPSS = 0.216 (p = 0.10) does not reject stationarity; Rolling two-sample KS (ref=120, win=120, step=6) mostly below KS_thr = 0.21; We trigger re-estimation if KS_thr is exceeded for q = 3 consecutive windows. Figures and a short report are written to out_1990/ (summary.md, rolling_cdf_ks.png). Data & preprocessing CPIAUCSL from FRED; Δlog×100; drop NAs; filter to 1990+. All CSVs are ASCII-only (no Unicode symbols). Contact / citation Please cite the SSRN preprint and this dataset; include the package SHA-256 above for verification.