The Cybersecurity Dataset Observatory: Which Data for AI in Cybersecurity?
Organizations: University of Napoli Federico II, Italy
Abstract
AI is increasingly used for cybersecurity tasks such as intrusion and malware detection, in enterprise, industrial, vehicular and mobile networks. Its effectiveness depends on data: models need representative network traffic, both to be trained and to be evaluated realistically. Public datasets of cybersecurity network traffic, however, are scattered across laboratory pages, repositories and aggregators, and are continuously released, corrected and withdrawn: 92 of the 234 datasets we catalog were released in 2023 or later. At this pace, surveys that list them are outdated soon after publication, since they describe the field only as it was when they were written. We present the Cybersecurity dataset observatory, a public and continuously updated catalog of cybersecurity datasets based on network traffic. New datasets are added as they are found or submitted, every record is periodically re-checked against its original source, to which it links rather than hosting the data, and every change is recorded in a public change log. This paper documents the first release: we define a traffic-centered scope and a seven-axis taxonomy, and review 234 datasets (204 original captures and 30 derivatives) across 14 domains. The review shows a field that grows quickly but unevenly, where 91 datasets distribute only processed records, a few laboratories produce a large share of the data, and UAV, wireless and V2X networks have little public evidence, while 5G has few real captures. We then relate the catalog to common uses in AI research, outline a procedure for selecting datasets, and discuss benchmark pitfalls and open gaps.
Figures & tables
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | |||||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | D | F | S | P | O | |||||
| OTIDS [ 144 ] | 2017 | ✓ | ✗ | – | ● | W | ✓ | 3 | ✓ | ✓ | ✓ | ✗ | ✗ | |
| HCRL Survival Analysis [ 103 ] | 2018 | ✓ | ✗ | – | ● | B | ✓ | 3 | ✓ | ✓ | ✓ | ✗ | ✗ | |
| HCRL Car-Hacking [ 209 ] | 2018 | ✓ | ✗ | – | ● | B | ✓ | 4 | ✓ | ✓ | ✓ | ✗ | ✗ | |
| TU/e CAN v2 [ 73 ] | 2019 | ✓ | ✗ | – | ◐ | W | ✓ | 5 | ✓ | ✓ | ✗ | ✓ | ✓ | |
| SynCAN [ 105 ] | 2019 | ✗ | ✓ | CAN-D | ○ | B | ✓ | 5 | ✓ | ✗ | ✓ | ✓ | ✓ | |
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | ||||||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | D | F | P | T | Y | O | |||||
| VeReMi [ 241 ] | 2018 | ✓ | ✗ | – | ○ | M | ✓ | 5 | ✗ | ✓ | ✗ | ✗ | ✗ | ✓ | |
| V2X Security Threats [ 93 ] | 2020 | ✓ | ✗ | – | ○ | M | – | – | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | |
| VeReMi Extension [ 125 ] | 2020 | ✓ | ✗ | – | ○ | M | ✓ | 19 | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | |
| VANET DDoS [ 253 ] | 2022 | ✓ | ✗ | – | ○ | – | ✓ | 4 | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | |
| MisbehaviorX [ 211 ] | 2024 | ✓ | ✗ | – | ○ | M | ✓ | 68 | – | ||||||
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | |||||||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | B | D | P | R | S | V | O | |||||
| UAV Cyber-Physical [ 58 ] | 2023 | ✗ | ✓ | CFM | ◐ | M | ✓ | 4 | ✗ | ✓ | ✓ | ✗ | ✓ | ✓ | ✗ | |
| ISOT Drone [ 118 ] | 2024 | ✓ | ✓ | custom | ◐ | B | ✓ | 5 | ✗ | ✗ | ✓ | ✗ | ✓ | ✓ | ✗ | |
| UAV-NIDD [ 59 ] | 2025 | ✓ | ✓ | CFM | ◐ | M | ✓ | 10 | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✓ | |
| UAVIDS-2025 [ 235 ] | 2025 | ✗ | ✓ | custom | ○ | M | ✓ | 4 | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✓ | |
| GCS-NIDD [ 60 ] | 2026 | ✓ | ✓ | CFM, Zeek | ◐ | M | ✓ | 9 | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✓ | |
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | |||||||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | B | D | F | P | R | S | O | |||||
| 5G-NIDD [ 205 ] | 2022 | ✓ | ✓ | Argus | ◐ | M | ✓ | 8 | ✗ | ✓ | ✗ | ✗ | ✓ | ✗ | ✗ | |
| 5GAD-2022 [ 54 ] | 2022 | ✓ | ✗ | – | ○ | M | ✓ | 10 | ✗ | ✗ | ✗ | ✗ | ✓ | ✓ | ✓ | |
| 5G Core IDS [ 215 ] | 2023 | ✗ | ✓ | tshark | ○ | B | ✓ | 3 | ✗ | ✓ | ✗ | ✗ | ✗ | ✓ | ✓ | |
| 5GC PFCP [ 11 ] | 2023 | ✓ | ✓ | CFM, custom | ○ | M | ✓ | 4 | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✓ | |
| 5G Slicing DoS/DDoS [ 131 ] | 2023 | ✓ | ✓ | CFM | ○ | – | ✓ | – | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | |
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | |||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | D | R | V | |||||
| DARPA 2000 LLDOS [ 164 ] | 2000 | ✓ | ✗ | – | ○ | W | – | 5 | ✓ | ✓ | ✓ | |
| CAIDA DDoS Attack 2007 [ 49 ] | 2007 | ✓ | ✗ | – | ● | – | ✗ | 1 | ✓ | ✗ | ✗ | |
| CIC-DDoS2019 [ 213 ] | 2019 | ✓ | ✓ | CFM | ◐ | M | ✓ | 13 | ✓ | ✓ | ✗ | |
| BOUN DDoS [ 76 ] | 2020 | ✗ | ✓ | Wireshark | ● | W | ✓ | 2 | ✓ | ✗ | ✗ | |
| USB-IDS-1 [ 47 ] | 2021 | ✓ | ✓ | CFM | ◐ | B | ✓ | 4 | ✓ | ✗ | ✗ | |
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | ||||||||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | B | C | D | M | R | V | W | O | |||||
| ISOT-CID [ 6 ] | 2018 | ✓ | ✗ | – | ● | B | ✓ | 1 | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | ✓ | ✓ | |
| SDN DDoS [ 3 ] | 2020 | ✗ | ✓ | Ryu | ○ | B | ✓ | 1 | ✗ | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | |
| InSDN [ 74 ] | 2020 | ✓ | ✓ | CFM | ○ | M | ✓ | 7 | ✓ | ✓ | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | |
| UNR-IDD [ 65 ] | 2023 | ✗ | ✓ | ONOS | ○ | M | ✓ | 5 | ✗ | ✗ | ✓ | ✗ | ✓ | ✗ | ✗ | ✓ | |
| BCCC-cPacket-Cloud-DDoS-2024 [ 25 ] | 2024 | ✓ | ✓ | NTLFlowLyzer | ◐ | M | ✓ | 17 | ✗ | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | |
| Dataset | Year | Artifact | Access | Acq. | Lab. | Classes | Attack families | |||||||||
| Raw | Proc. | Proc. Tool | Ben. | Att. | B | C | M | R | F | G | O | |||||
| USTC-TFC2016 [ 245 ] | 2016 | ✓ | ✗ | – | mix | W | ✓ | 10 | ✗ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | |
| HIKARI-2021 [ 79 ] | 2021 | ✓ | ✓ | Zeek | ◐ | M | ✓ | 4 | ✓ | ✗ | ✓ | ✓ | ✗ | ✗ | ✗ | |
| CIC-PQC_OAV v1 [ 43 ] | 2025 | ✗ | ✓ | tshark | ○ | M | ✓ | 1 | ✗ | ✗ | ✗ | ✗ | ✓ | ✓ | ✓ | |
| TQH-C2 [ 233 ] | 2026 | ✓ | ✓ | Zeek, JA4 | ○ | M | – | – | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | |
| Use | Required | Related columns | Caution | Datasets | |
| Supervised classification | Lab. = B, M or ML | Ben., Att., Attack families | Weak labels (W), assigned per file, time interval or IDS alert, are coarse or noisy; class definitions differ across datasets, so multiclass results are not directly comparable | 184 | |
| Anomaly detection | Ben. = ✓ | Acq. | Testbed benign traffic comes from few, often scripted hosts and inflates detection rates; benign traffic from real networks may contain undetected attacks | 201 | |
| Learning setting | Self-supervised pretraining | Raw = ✓ | Acq., lineage ( 2 ) | Pretraining on a capture and testing on one of its derivatives ( 2 ) leaks test data | 141 |
| Packet-level and sequence models on IP traffic | Raw = ✓, domain {automotive, V2X} | Proc. Tool, Lab. | Labels are often given per flow in the processed records and must be mapped back to packets | 113 | |
| Input representation | CAN and V2X models | Domain = {automotive or V2X} | Raw, Acq., Lab. | CAN identifiers and payloads are vehicle-specific and attacks are injected; all V2X datasets are simulated | 30 |
| Case 1: IoT intrusion detection | Case 2: 5G realism | Case 3: ICS gaps | ||||
| Step | Filter | # | Filter | # | Filter | # |
| 1 Domain | IoT / IIoT / IoMT | 50 | 5G / mobile core | 13 | ICS / OT | 22 |
| 2 Task | Packet-level, supervised | Supervised | Supervised | |||
| 3 Representation | Raw traces | 32 | Raw traces | 9 | Raw traces | 15 |
| 4 Labels | Multiclass | 24 | Multiclass | 5 | Multiclass | 7 |
| Benign class | 19 | – | – | |||
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
| Dataset | Domain | Year | Raw | Proc. | Acq. | Lab. | Att. | Status | Evidence |
| CDX CTF 2009 [ 206 ] | Enterprise / general IP | 2009 | ✓ | ✗ | ◐ | – | – | Withdrawn | The usma.edu pages are offline after the move to westpoint.edu; derived features remain available in ASNM-CDX-2009. |
| TUIDS [ 89 ] | DDoS | 2012 | ✓ | ✓ | ◐ | B | 8 | Withdrawn | The paper points to a Tezpur University page that is no longer online; listed as unavailable by COMIDDS. |
| ISCX Botnet 2014 [ 34 ] | Malware / botnet / C2 | 2014 | ✓ | ✗ | ◐ | B | 16 | Withdrawn | Confirmed as unavailable on the official UNB/CIC download platform. |
| SSHCure [ 110 ] | Backbone / ISP / Honeypot | 2014 | ✗ | ✓ | ● | – | – | Withdrawn | SimpleWeb lists the flow and log files of the University of Twente campus network, but no public download link is available. |
| CIC DoS 2017 [ 120 ] | DDoS | 2017 | ✓ | ✗ | ◐ | – | 8 | Withdrawn | Confirmed as currently unavailable on the official UNB/CIC download page. |
| WSN-BFSF [ 68 ] | Wireless / LPWAN / WSN | 2024 | ✗ | ✓ | ○ | – | 3 | Data not published | The paper (NS-2 simulation, 312k records) exists; no official download location was found. |
| Acronym | Meaning | Acronym | Meaning |
| 5GC | 5G Core network | LoRaWAN | Long Range Wide-Area Network |
| AI | Artificial Intelligence | LPWAN | Low-Power Wide-Area Network |
| AMF | Access and Mobility Management Function (5G core) | M2M | Machine-to-Machine |
| AP | Access Point | MAC | Medium Access Control |
| APT | Advanced Persistent Threat | MAWI | Measurement and Analysis on the WIDE Internet |
| ATT&CK | Adversarial Tactics, Techniques, and Common Knowledge (MITRE) | MitM | Man-in-the-Middle |