keys-i commited on
Commit
59b97b2
·
verified ·
1 Parent(s): c530c3a

data: publish verified NetFlow V3 Parquet

Browse files

Reorganize the four UQ NetFlow V3 Parquet files under data/, move the supplied feature dictionary under metadata/, and replace the dataset card with measured row counts, source DOIs, conversion checks, schema differences, leakage warnings, and corrected UQ Open Access terms.

README.md CHANGED
@@ -1,57 +1,148 @@
1
  ---
2
- license: cc-by-nc-sa-4.0
3
  pretty_name: NetFlow V3
 
 
 
4
  language:
5
  - en
6
  tags:
7
- - IDS
8
- - NIDS
9
- - NetFlow
10
- - parquet
11
  - cybersecurity
 
 
12
  size_categories:
13
  - 10M<n<100M
 
 
14
  configs:
15
- - config_name: NF-CICIDS2018
 
16
  data_files:
17
- - split: train
18
- path: NF-CICIDS2018-v3.parquet
19
-
20
- - config_name: NF-BoT-IoT
21
  data_files:
22
- - split: train
23
- path: NF-BoT-IoT-v3.parquet
24
-
25
- - config_name: NF-ToN-IoT
26
  data_files:
27
- - split: train
28
- path: NF-ToN-IoT-v3.parquet
29
-
30
- - config_name: NF-UNSW-NB15
31
  data_files:
32
- - split: train
33
- path: NF-UNSW-NB15-v3.parquet
34
  ---
35
 
36
  # NetFlow V3
37
 
38
- NetFlow V3 datasets for machine learning-based network intrusion detection research.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39
 
40
- This repository republishes the NetFlow V3 datasets introduced in *Temporal Analysis of NetFlow Datasets for Network Intrusion Detection Systems* by Majed Luay, Siamak Layeghy, Seyedehfaezeh Hosseininoorbin, Mohanad Sarhan, Nour Moustafa, and Marius Portmann. [arXiv](https://arxiv.org/abs/2503.04404)
41
 
42
- ## Description
 
 
 
 
 
 
 
 
 
 
 
 
43
 
44
- NetFlow V3 extends earlier NetFlow datasets with temporal features for time-based analysis in NIDS research. [arXiv](https://arxiv.org/abs/2503.04404)
45
 
46
- ## Citation
 
 
 
 
 
 
47
 
48
- If you use this dataset, please cite the original paper:
 
49
 
50
  ```bibtex
51
- @article{luay2025temporal,
52
- title={Temporal Analysis of NetFlow Datasets for Network Intrusion Detection Systems},
53
- author={Luay, Majed and Layeghy, Siamak and Hosseininoorbin, Seyedehfaezeh and Sarhan, Mohanad and Moustafa, Nour and Portmann, Marius},
54
- journal={arXiv preprint arXiv:2503.04404},
55
- year={2025},
56
- doi={10.48550/arXiv.2503.04404}
57
- }
 
 
 
 
 
 
 
1
  ---
 
2
  pretty_name: NetFlow V3
3
+ license: other
4
+ license_name: uq-open-access-data-terms
5
+ license_link: https://web.library.uq.edu.au/research-and-publish/manage-research-data/share-and-reuse-research-data
6
  language:
7
  - en
8
  tags:
9
+ - tabular
10
+ - timeseries
 
 
11
  - cybersecurity
12
+ - netflow
13
+ - network-intrusion-detection
14
  size_categories:
15
  - 10M<n<100M
16
+ task_categories:
17
+ - tabular-classification
18
  configs:
19
+ - config_name: NF-UNSW-NB15-v3
20
+ default: true
21
  data_files:
22
+ - split: full
23
+ path: data/NF-UNSW-NB15-v3.parquet
24
+ - config_name: NF-BoT-IoT-v3
 
25
  data_files:
26
+ - split: full
27
+ path: data/NF-BoT-IoT-v3.parquet
28
+ - config_name: NF-ToN-IoT-v3
 
29
  data_files:
30
+ - split: full
31
+ path: data/NF-ToN-IoT-v3.parquet
32
+ - config_name: NF-CSE-CIC-IDS2018-v3
 
33
  data_files:
34
+ - split: full
35
+ path: data/NF-CICIDS2018-v3.parquet
36
  ---
37
 
38
  # NetFlow V3
39
 
40
+ Parquet mirror of the four NetFlow V3 datasets published by the University of
41
+ Queensland. The files contain 53 flow and temporal features plus the binary
42
+ `Label` and multiclass `Attack` targets. Together they contain 66,935,021
43
+ completed bidirectional flow records.
44
+
45
+ ## Contents
46
+
47
+ | Configuration | Rows | Attack rows | `Attack` values | Observed flow period (UTC) | Official record |
48
+ |---|---:|---:|---:|---|---|
49
+ | `NF-UNSW-NB15-v3` | 2,365,424 | 127,693 | 10 | 2015-01-22 to 2015-02-18 | [10.48610/6E0EDA1](https://doi.org/10.48610/6E0EDA1) |
50
+ | `NF-BoT-IoT-v3` | 16,933,808 | 16,881,819 | 5 | 2018-05-15 to 2018-06-19 | [10.48610/73C4EBC](https://doi.org/10.48610/73C4EBC) |
51
+ | `NF-ToN-IoT-v3` | 27,520,260 | 10,728,046 | 10 | 2019-04-23 to 2019-04-29 | [10.48610/44D7C5E](https://doi.org/10.48610/44D7C5E) |
52
+ | `NF-CSE-CIC-IDS2018-v3` | 20,115,529 | 2,600,903 | 15 | 2018-02-14 to 2018-03-02 | [10.48610/ECE9B83](https://doi.org/10.48610/ECE9B83) |
53
+
54
+ The source archive names the final CSV `NF-CICIDS2018-v3.csv`; the Hub
55
+ configuration retains the official `NF-CSE-CIC-IDS2018-v3` dataset name.
56
+ `metadata/NetFlow_v3_Features.csv` is the feature dictionary supplied by UQ.
57
+
58
+ ## Loading
59
+
60
+ ```python
61
+ from datasets import load_dataset
62
+
63
+ flows = load_dataset(
64
+ "keys-i/netFlow",
65
+ "NF-UNSW-NB15-v3",
66
+ split="full",
67
+ )
68
+ ```
69
+
70
+ `full` is a storage split only. UQ does not provide one common leakage-safe
71
+ train/validation/test split for these four files. Do not interpret it as a
72
+ training partition.
73
+
74
+ ## Conversion and checks
75
+
76
+ The source packages were downloaded from the four UQ records above. Their
77
+ BagIt payload SHA-1 checks passed before conversion. Each CSV was converted
78
+ with Zstandard compression using
79
+ [`tools/scripts/convert.py`](https://github.com/keys-i/FlowIDS/blob/main/tools/scripts/convert.py):
80
+
81
+ ```console
82
+ pixi run python tools/scripts/convert.py \
83
+ --input <source.csv> \
84
+ --output <dataset.parquet> \
85
+ --overwrite
86
+ ```
87
+
88
+ No filtering, relabelling, deduplication, or feature engineering is performed
89
+ by that converter. Verification on 2026-08-16 established that every Parquet
90
+ row count equals its source CSV line count minus the header, all files have the
91
+ same 55 column names in the same order, and every file opens as Zstandard
92
+ Parquet.
93
+
94
+ Three inferred numeric types differ across configurations because the source
95
+ values differ:
96
+
97
+ | Field | UNSW / BoT-IoT | ToN-IoT / CSE-CIC-IDS2018 |
98
+ |---|---|---|
99
+ | `L7_PROTO` | `float64` | `int64` |
100
+ | `SRC_TO_DST_SECOND_BYTES` | `float64` | `int64` |
101
+ | `DST_TO_SRC_SECOND_BYTES` | `float64` | `int64` |
102
+
103
+ Cast these fields explicitly before concatenating configurations. Do not rely
104
+ on an implicit cross-dataset coercion.
105
 
106
+ ## Research-use limits
107
 
108
+ - These are converted benchmark lineages, not four independent operational
109
+ networks collected under one protocol.
110
+ - The files include endpoint addresses, absolute timestamps, labels, and other
111
+ fields that can leak capture or scenario identity. A random row split does
112
+ not demonstrate temporal, endpoint, attack-family, or cross-network
113
+ generalisation.
114
+ - `Label` and `Attack` must be hidden from self-supervised pretraining. Dataset
115
+ taxonomies are configuration-specific and must not be merged by fuzzy name
116
+ matching.
117
+ - Duplicate and near-duplicate flows must be audited before constructing any
118
+ evaluation split.
119
+ - Results on these four files alone are benchmark evidence, not proof of a
120
+ general-purpose network foundation model or operational intrusion detector.
121
 
122
+ ## Rights and attribution
123
 
124
+ Each UQ catalogue record reports open access and lists its licence as “Other.”
125
+ UQ currently defines open access research data as available for access, use,
126
+ reuse, and redistribution. This mirror therefore uses Hugging Face's `other`
127
+ licence marker, links to that UQ guidance, and does not relicense the source
128
+ data. Review the applicable UQ record and terms before reuse. The previous
129
+ dataset card's CC-BY-NC-SA-4.0 label was not supported by the UQ records and
130
+ has been removed.
131
 
132
+ Please cite the applicable dataset DOI from the contents table and the V3
133
+ feature paper:
134
 
135
  ```bibtex
136
+ @article{luay2026time,
137
+ author = {Majed Luay and Siamak Layeghy and Niloufar Noorbin and
138
+ Mohanad Sarhan and Gayan Kulatilleke and Nour Moustafa and
139
+ Marius Portmann},
140
+ title = {Time Matters: Temporal NetFlow Features for ML-Based Network
141
+ Intrusion Detection},
142
+ journal = {IEEE Access},
143
+ volume = {14},
144
+ pages = {66899--66913},
145
+ year = {2026},
146
+ doi = {10.1109/ACCESS.2026.3688204}
147
+ }
148
+ ```
NF-BoT-IoT-v3.parquet → data/NF-BoT-IoT-v3.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0679550f7b6cd09d487f5c78f4b8e7f13c7fcd5e70970210b06e1e5295b9c13e
3
- size 63896742
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4228c33fea2c599d572d92e4d25256153e1113b8b1ac2aea01ff236144a7ae34
3
+ size 70270946
NF-CICIDS2018-v3.parquet → data/NF-CICIDS2018-v3.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cd4f64d381eebe56029876868bc7dc1f96f4e681b38d154c9a465b0388b51636
3
- size 809943210
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80f67e99c71808bfe57b34932fa35c9eaff8dbb856b5a0aefb9d278ee81230ea
3
+ size 815742559
NF-ToN-IoT-v3.parquet → data/NF-ToN-IoT-v3.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3b4bb7506c9a2a56f3782cab3f9ca9bd67fbda2e6950418943148a7bfdd6bb53
3
- size 442356435
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:074edcc2cf1e75b7dc7098acd628525d4c0a23e9d60ed43ff379edddf31c8a21
3
+ size 450120816
NF-UNSW-NB15-v3.parquet → data/NF-UNSW-NB15-v3.parquet RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c3ac1d4d4d0ff7f199e1db36886f18e36567f5fc464051fdfb5ecbd94842e713
3
- size 103828433
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:780ae4846881a047d1c8ddb08cdbae17ab9dcb8acbbd42982e45bab07beb8d59
3
+ size 105425657
NetFlow_v3_Features.csv → metadata/NetFlow_v3_Features.csv RENAMED
@@ -1,54 +1,54 @@
1
- Feature,Description
2
- IPV4_SRC_ADDR,IPv4 source address
3
- IPV4_DST_ADDR,IPv4 destination address
4
- L4_SRC_PORT,IPv4 source port number
5
- L4_DST_PORT,IPv4 destination port number
6
- PROTOCOL,IP protocol identifier byte
7
- L7_PROTO,Layer 7 protocol (numeric)
8
- IN_BYTES,Incoming number of bytes
9
- OUT_BYTES,Outgoing number of bytes
10
- IN_PKTS,Incoming number of packets
11
- OUT_PKTS,Outgoing number of packets
12
- FLOW_DURATION_MILLISECONDS,Flow duration in milliseconds
13
- TCP_FLAGS,Cumulative of all TCP flags
14
- CLIENT_TCP_FLAGS,Cumulative of all client TCP flags
15
- SERVER_TCP_FLAGS,Cumulative of all server TCP flags
16
- DURATION_IN,Client to Server stream duration (msec)
17
- DURATION_OUT,Client to Server stream duration (msec)
18
- MIN_TTL,Min flow TTL
19
- MAX_TTL,Max flow TTL
20
- LONGEST_FLOW_PKT,Longest packet (bytes) of the flow
21
- SHORTEST_FLOW_PKT,Shortest packet (bytes) of the flow
22
- MIN_IP_PKT_LEN,Len of the smallest flow IP packet observed
23
- MAX_IP_PKT_LEN,Len of the largest flow IP packet observed
24
- SRC_TO_DST_SECOND_BYTES,Src to dst Bytes/sec
25
- DST_TO_SRC_SECOND_BYTES,Dst to src Bytes/sec
26
- RETRANSMITTED_IN_BYTES,Number of retransmitted TCP flow bytes (src->dst)
27
- RETRANSMITTED_IN_PKTS,Number of retransmitted TCP flow packets (src->dst)
28
- RETRANSMITTED_OUT_BYTES,Number of retransmitted TCP flow bytes (dst->src)
29
- RETRANSMITTED_OUT_PKTS,Number of retransmitted TCP flow packets (dst->src)
30
- SRC_TO_DST_AVG_THROUGHPUT,Src to dst average thpt (bps)
31
- DST_TO_SRC_AVG_THROUGHPUT,Dst to src average thpt (bps)
32
- NUM_PKTS_UP_TO_128_BYTES,Packets whose IP size <= 128
33
- NUM_PKTS_128_TO_256_BYTES,Packets whose IP size > 128 and <= 256
34
- NUM_PKTS_256_TO_512_BYTES,Packets whose IP size > 256 and <= 512
35
- NUM_PKTS_512_TO_1024_BYTES,Packets whose IP size > 512 and <= 1024
36
- NUM_PKTS_1024_TO_1514_BYTES,Packets whose IP size >��1024 and <= 1514
37
- TCP_WIN_MAX_IN,Max TCP Window (src->dst)
38
- TCP_WIN_MAX_OUT,Max TCP Window (dst->src)
39
- ICMP_TYPE,ICMP Type * 256 + ICMP code
40
- ICMP_IPV4_TYPE,ICMP Type
41
- DNS_QUERY_ID,DNS query transaction Id
42
- DNS_QUERY_TYPE,"DNS query type (e.g. 1=A, 2=NS..)"
43
- DNS_TTL_ANSWER,TTL of the first A record (if any)
44
- FTP_COMMAND_RET_CODE,FTP client command return code
45
- FLOW_START_MILLISECONDS,Flow start timestamp in milliseconds
46
- FLOW_END_MILLISECONDS,Flow end timestamp in milliseconds
47
- SRC_TO_DST_IAT_MIN ,Minimum Inter-Packet Arrval Time (src->dst)
48
- SRC_TO_DST_IAT_MAX ,Maximum Inter-Packet Arrval Time (src->dst)
49
- SRC_TO_DST_IAT_AVG ,Average Inter-Packet Arrval Time (src->dst)
50
- SRC_TO_DST_IAT_STDDEV ,Sandard Deviaion Inter-Packet Arrval Time (src->dst)
51
- DST_TO_SRC_IAT_MIN ,Minimum Inter-Packet Arrval Time (dst > src)
52
- DST_TO_SRC_IAT_MAX ,Minimum Inter-Packet Arrval Time (dst > src)
53
- DST_TO_SRC_IAT_AVG ,Minimum Inter-Packet Arrval Time (dst > src)
54
- DST_TO_SRC_IAT_STDDEV ,Minimum Inter-Packet Arrval Time (dst > src)
 
1
+ Feature,Description
2
+ IPV4_SRC_ADDR,IPv4 source address
3
+ IPV4_DST_ADDR,IPv4 destination address
4
+ L4_SRC_PORT,IPv4 source port number
5
+ L4_DST_PORT,IPv4 destination port number
6
+ PROTOCOL,IP protocol identifier byte
7
+ L7_PROTO,Layer 7 protocol (numeric)
8
+ IN_BYTES,Incoming number of bytes
9
+ OUT_BYTES,Outgoing number of bytes
10
+ IN_PKTS,Incoming number of packets
11
+ OUT_PKTS,Outgoing number of packets
12
+ FLOW_DURATION_MILLISECONDS,Flow duration in milliseconds
13
+ TCP_FLAGS,Cumulative of all TCP flags
14
+ CLIENT_TCP_FLAGS,Cumulative of all client TCP flags
15
+ SERVER_TCP_FLAGS,Cumulative of all server TCP flags
16
+ DURATION_IN,Client to Server stream duration (msec)
17
+ DURATION_OUT,Client to Server stream duration (msec)
18
+ MIN_TTL,Min flow TTL
19
+ MAX_TTL,Max flow TTL
20
+ LONGEST_FLOW_PKT,Longest packet (bytes) of the flow
21
+ SHORTEST_FLOW_PKT,Shortest packet (bytes) of the flow
22
+ MIN_IP_PKT_LEN,Len of the smallest flow IP packet observed
23
+ MAX_IP_PKT_LEN,Len of the largest flow IP packet observed
24
+ SRC_TO_DST_SECOND_BYTES,Src to dst Bytes/sec
25
+ DST_TO_SRC_SECOND_BYTES,Dst to src Bytes/sec
26
+ RETRANSMITTED_IN_BYTES,Number of retransmitted TCP flow bytes (src->dst)
27
+ RETRANSMITTED_IN_PKTS,Number of retransmitted TCP flow packets (src->dst)
28
+ RETRANSMITTED_OUT_BYTES,Number of retransmitted TCP flow bytes (dst->src)
29
+ RETRANSMITTED_OUT_PKTS,Number of retransmitted TCP flow packets (dst->src)
30
+ SRC_TO_DST_AVG_THROUGHPUT,Src to dst average thpt (bps)
31
+ DST_TO_SRC_AVG_THROUGHPUT,Dst to src average thpt (bps)
32
+ NUM_PKTS_UP_TO_128_BYTES,Packets whose IP size <= 128
33
+ NUM_PKTS_128_TO_256_BYTES,Packets whose IP size > 128 and <= 256
34
+ NUM_PKTS_256_TO_512_BYTES,Packets whose IP size > 256 and <= 512
35
+ NUM_PKTS_512_TO_1024_BYTES,Packets whose IP size > 512 and <= 1024
36
+ NUM_PKTS_1024_TO_1514_BYTES,Packets whose IP size >��1024 and <= 1514
37
+ TCP_WIN_MAX_IN,Max TCP Window (src->dst)
38
+ TCP_WIN_MAX_OUT,Max TCP Window (dst->src)
39
+ ICMP_TYPE,ICMP Type * 256 + ICMP code
40
+ ICMP_IPV4_TYPE,ICMP Type
41
+ DNS_QUERY_ID,DNS query transaction Id
42
+ DNS_QUERY_TYPE,"DNS query type (e.g. 1=A, 2=NS..)"
43
+ DNS_TTL_ANSWER,TTL of the first A record (if any)
44
+ FTP_COMMAND_RET_CODE,FTP client command return code
45
+ FLOW_START_MILLISECONDS,Flow start timestamp in milliseconds
46
+ FLOW_END_MILLISECONDS,Flow end timestamp in milliseconds
47
+ SRC_TO_DST_IAT_MIN ,Minimum Inter-Packet Arrval Time (src->dst)
48
+ SRC_TO_DST_IAT_MAX ,Maximum Inter-Packet Arrval Time (src->dst)
49
+ SRC_TO_DST_IAT_AVG ,Average Inter-Packet Arrval Time (src->dst)
50
+ SRC_TO_DST_IAT_STDDEV ,Sandard Deviaion Inter-Packet Arrval Time (src->dst)
51
+ DST_TO_SRC_IAT_MIN ,Minimum Inter-Packet Arrval Time (dst > src)
52
+ DST_TO_SRC_IAT_MAX ,Minimum Inter-Packet Arrval Time (dst > src)
53
+ DST_TO_SRC_IAT_AVG ,Minimum Inter-Packet Arrval Time (dst > src)
54
+ DST_TO_SRC_IAT_STDDEV ,Minimum Inter-Packet Arrval Time (dst > src)