Descriptor: Multi-Regional Cloud Honeynet Dataset (MURHCAD)
Enrique Feito-Casares, Ismael Gómez-Talal, José-Luis Rojo-Álvarez · 2026
Read Descriptor: Multi-Regional Cloud Honeynet Dataset (MURHCAD) by Enrique Feito-Casares, Ismael Gómez-Talal, José-Luis Rojo-Álvarez (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-05813.txt.
Introduction
Enrique Feito-Casares and 2 co-authors describe in 2026 a honeypot dataset capturing three days of global cyberattacks against cloud decoy systems. Honeypots are machines left exposed on purpose to record attacker behavior. Their network ran three honeypot programs on four Azure virtual machines in different regions during June 2025.
The dataset holds over one hundred thirty thousand attack events with timestamps, addresses, geolocation, targeted ports, and protocol labels. A small set of source addresses spanned ninety-five countries, while three protocols dominated the traffic. Activity peaked at certain hours, and each honeypot attracted geographically distinct attack mixes. The authors provide analysis code and guidance for using the data in anomaly detection and threat intelligence work.
Catalog record
| CREATOR | Enrique Feito-Casares, Ismael Gómez-Talal, José-Luis Rojo-Álvarez |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-05813
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Descriptor: Multi-Regional Cloud Honeynet Dataset (MURHCAD) Thanks: This work has been submitted to the IEEE for possible publication Enrique Feito-Casares, Ismael Gómez-Talal, José-Luis Rojo-Álvarez Abstract This data article introduces a comprehensive, high-resolution honeynet dataset designed to support standalone analyses of global cyberattack behaviors. Collected over a continuous 72-hour window (June 9 to 11, 2025) on Microsoft Azure, the dataset comprises 132,425 individual attack events captured by three honeypots (Cowrie, Dionaea, and SentryPeer) deployed across four geographically dispersed virtual machines. Each event record includes enriched metadata (UTC timestamps, source/destination IPs, autonomous system and organizational mappings, geolocation coordinates, targeted ports, and honeypot identifiers alongside derived temporal features and standardized protocol classifications). We provide actionable guidance for researchers seeking to leverage this dataset in anomaly detection, protocol-misuse studies, threat intelligence, and defensive policy design. Descriptive statistics highlight significant skew: 2,438 unique source IPs span 95 countries, yet the top 1% of IPs account for 15% of all events, and three protocols dominate: Session Initiation Protocol (SIP), Telnet, Server Message Block (SMB). Temporal analysis uncovers pronounced rush-hour peaks at 07:00 and 23:00 UTC, interspersed with maintenance-induced gaps that reveal operational blind spots. Geospatial mapping further underscores platform-specific biases: SentryPeer captures concentrated SIP floods in North America and Southeast Asia, Cowrie logs Telnet/SSH scans predominantly from Western Europe and the U.S., and Dionaea records SMB exploits around European nodes. By combining fine-grained temporal resolution with rich, contextual geolocation and protocol metadata, this standalone dataset aims to empower reproducible, cloud-scale investigations into evolving cyber threats. Accompanying analysis code and data access details are provided to facilitate immediate adoption and foster further methodological innovations in cybersecurity research. Keywords Anomaly Detection \cdot Azure Cloud Deployment \cdot Cyberattack Dataset \cdot Honeynet \cdot Temporal Attack Patterns BACKGROUND The design and evaluation of effective cybersecurity solutions strongly depend on the availability of realistic and diverse datasets. Traditional datasets have frequently been limited either in scope or realism, often lacking real-time attack traces or contextual metadata. Recent contributions aim to address this gap by providing curated traffic traces, logs, and behavioral patterns in controlled but realistic environments. For example, a dataset was proposed for capturing comprehensive attack scenarios against vulnerable virtual machines, including FTP and SSH exploitation, ARP spoofing, and brute-force attempts, all observed via tools such as Wireshark and tcpdump ( 11 ) . Cyber-physical systems have been explored through labeled datasets with breakdowns, anomalies, and cyberattacks on fluid control subsystems, providing valuable insight into security monitoring in industrial contexts ( 6 ) . Dynamic and large-scale initiatives have contributed realistic network traffic and system logs obtained during multi-day cyber defense exercises ( 13 ) , allowing researchers to test anomaly detection methods under simulated attack–defense scenarios. At the TCP level, buffering delays in cyber-physical systems and Internet of Things (IoT) environments have been characterized, offering fine-grained data on how protocol-level interactions can degrade responsiveness in real-time systems ( 1 ) . In the IoT domain, where device heterogeneity and Distributed Denial of Service (DDoS) threats pose unique challenges, recent datasets focus on classifying and detecting diverse attack types in heterogeneous networks, such as IoT-DH ( 9 ) . Large-scale industrial data from optical–wireless sensor networks has also been made available, enabling cyber-physical systems performance analysis with metrics such as latency, throughput, and packet error rates ( 3 ) . More relevant to our work are honeypot-based initiatives. Among these, Hornet 40 ( 14 ) and Hornet 65 Niner ( 15 ) emphasize the role of geographical location in attack exposure, aggregating millions of flow records from passive sensors and honeypots deployed globally (e.g., New York, London, Bangalore), and providing standardized Zeek or Argus logs. To address this gap, we propose the MURHCAD , a high-resolution collection of inbound cyberattacks captured over a 72-hour period via a globally distributed honeynet on Microsoft Azure. This dataset facilitates novel investigations into attack bursts, diurnal rhythms, protocol-level misuse, and concentration of malicious behavior by source IPs and organizations. In addition to canonical attributes such as timestamp, protocol, and port information, the dataset enriches each event with source geolocation, organizational attribution (via IP-to-autonomous system number (ASN) mappings), and the targeted host. Unlike prior honeypot collections, this dataset places particular emphasis on temporal granularity-allowing researchers to investigate attack behavior patterns at the hourly level-and supports analysis of protocol misuse, attack bursts, and spatial-temporal attack dynamics. While previous honeypot datasets have offered valuable insights into global attack surfaces, they often lack high-resolution temporal labeling, detailed metadata on attacker provenance, and synchronized data across heterogeneous platforms. In particular, the literature still lacks a dataset that combines large-scale, geographically distributed honeypot traces with hourly-level resolution and enriched contextual metadata, enabling fine-grained spatiotemporal analyses ( 8 ) . The presented dataset ( 4 ) delivers a comprehensive portrait of real-world cyberattacks by combining honeypot logs from three distinct platforms (Cowrie, Dionaea, and SentryPeer 1 1 1 Registered trademark. ) deployed across four geographically dispersed virtual machines. A honeypot is a security mechanism designed to attract and study cyberattacks by simulating vulnerable systems. This dataset fine-grained temporal resolution enables researchers to dissect attack dynamics on an hourly basis, revealing characteristic daily rush hours and maintenance-induced blind spots. Simultaneously, protocol-level annotations (e.g., Session Initiation Protocol (SIP), Telnet, or Server Message Block (SMB)) highlight adversaries’ shifting focus among VoIP, legacy remote-login services, and Windows file-sharing endpoints, providing actionable insights for network segmentation and firewall policy design. Beyond temporal and protocol analyses, the Multi-Regional Cloud Honeynet Dataset offers rich geolocation metadata that supports cross-regional threat intelligence: over 2 400 unique source IPs spanning 95 countries illuminate global scanning patterns and concentrate attention on high-connectivity hubs. By mapping each intrusion to its autonomous system organization and country of origin, one can evaluate the effectiveness of anomaly-detection algorithms in distinguishing benign background noise from orchestrated scanning campaigns and further refine machine-learning classifiers for multi-class, time-series scenarios. Because this corpus captures both qualitative (attackType, protocol, srcCountry) and quantitative (timestamps, ports, latitude/longitude) dimensions within a consistent three-day window, it is especially well suited for benchmarking defensive mechanisms under controlled yet realistic conditions. Industrial practitioners may leverage these insights to optimize honeypot placement and adaptive monitoring, while academic researchers can use the data to develop and validate generative models, forecasting tools, and automated segmentation strategies that anticipate and mitigate emerging cyber threats. COLLECTION METHODS AND DESIGN Figure 1: Global Azure-based hub-and-spoke honeynet architecture containing four exposed sensor virtual machines, each within its virtual network, all peered to a central honeynet hub virtual network hosting the hive virtual machine and a VPN Gateway. The experimental protocol involved deploying four virtual machines across different geographic regions using the Azure cloud platform to mimic a realistic environment. Each virtual machine ran honeypot services to monitor and log activity. The virtual machines remained active from June 9 to June 11, enabling comprehensive data collection. The dataset described in this work was extracted using a dedicated infrastructure on the Microsoft Azure cloud provider and using the T-Pot Community Edition ( 12 ) . T-Pot Community Edition is an all-in-one, optionally distributed, multi-architecture honeypot platform supporting over 20 honeypots and offering rich visualization, animated live attack maps, and numerous security tools to enhance the deception experience. To capture high-value threat activity on key service ports,