The UCSD Network Telescope consists of a globally routed, but lightly utilized /8 network prefix, that is, 1/256th of the whole IPv4 address space. It contains few legitimate hosts; inbound traffic to non-existent machines - so called Internet Background Radiation (IBR) - is unsolicited and results from a wide range of events, including misconfiguration (e.g. mistyping an IP address), scanning of address space by attackers or malware looking for vulnerable targets, backscatter from randomly spoofed denial-of-service attacks, and the automated spread of malware. CAIDA continously captures this anomalous traffic discarding the legitimate traffic packets destined to the few reachable IP addresses in this prefix. We archive and aggregate these data, and provide this valuable resource to network security researchers.
Raw data captured by the UCSD Network Telescope are stored in huge pcap files, each file containing 1 hour of data. In order to enable more efficient data storage, processing, and analysis, these hourly pcap files are post-processed using Corsaro software to extract the most important packet header fields and aggregate data into FlowTuple files. The FlowTuple format includes the following eight fields:source IP address; destination IP address; source port; destination port; protocol; TCP Flags; TTL; IP length.In the hourly FlowTuple output files, the data are broken into 60-second intervals. Within a given interval, each unique key (a unique combination of the FlowTuple fields) observed in the raw pcap data is reported on a separate line in the following format:
where "value" is the number of packets in this interval whose header fields match this FlowTuple key.
Flows are further subdivided into three FlowTuple classes: backscatter, ICMP Request and "other" (not backscatter and not ICMP Request), and the total number of flows in each class is recorded.
One can use "cors2ascii" command to display the FlowTuple output in a human-readable ASCII format:
$ cors2ascii example.flowtuple.cors.gz #CORSARO_INTERVAL_START 0 1527811200 START flowtuple_backscatter 1012004 xxx.181.55.12|xx.68.45.71|3|3|1|19|0x00|100,20 xxx.85.229.197|xx.4.59.12|3|1|1|19|0x00|159,1 ...... END flowtuple_backscatter START flowtuple_icmpreq 314399 xxx.136.34.2|xx.128.122.196|8|0|1|1|0x00|44,3 xxx.52.83.2|xx.165.41.187|8|0|1|1|0x00|48,2 ...... END flowtuple_icmpreq START flowtuple_other 17964239 xxx.228.34.84|xx.151.31.6|0|0|4|63|0x00|256,6 xxx.228.34.84|xx.151.31.6|0|0|4|63|0x00|512,140 ...... END flowtuple_other # CORSARO_INTERVAL_END 0 1527811259 # CORSARO_INTERVAL_START 1 1527811260 START flowtuple_backscatter 991397 .........
Each hourly flowtuple file contains 60 intervals (0-59) characterized by unique start and end local unix time. More information about the structure of Corsaro global output file can be found at Corsaro Documentation page. This ongoing dataset covering the period from February 2008 till now is stored locally at CAIDA.
Caveats that apply to this dataset
This dataset and the types of worm and denial-of-service attack traffic contained therein are representative only of some spoofed source denial-of-service attacks. Many denial-of-service attackers do not spoof source IP addresses when they attack their victim, in which case backscatter would not appear on a telescope. Attackers can also spoof in a non-random fashion, which will incur an uneven distribution of backscatter across the IPv4 address space, and may cause backscatter traffic to miss any telescope lenses. Note that the telescope does not send any packets in response, which also limits insight into the traffic it sees.
Data Access Policy
These data must be analyzed on CAIDA machines, and cannot be downloaded!
Academic researchers and US government agencies can request access through CAIDA by filling out and submitting the online form. It usually takes about five to ten business days to process your request. We carefully review each application and the decision to grant the data access is based on the merits of your proposed data use.
These data also may be available for corporate entities who participate in CAIDA's membership program. Information on membership levels, services, and rates can be requested by emailing firstname.lastname@example.org.
Once users are approved for access to this dataset, they will receive an account on the CAIDA machine that provides direct access to the Telescope data they requested. Accounts are valid for a nominal twelve months in which the research is expected to be completed. CAIDA strictly enforces a "take software to the data" policy for this dataset: all analysis must be performed on CAIDA computers; download of raw data is not allowed. CAIDA provides several basic tools to work with the dataset, including CoralReef and Corsaro. Researchers can also upload their own analysis software.
Acceptable Use Agreement
Access to these data is subject to the terms of the following CAIDA Acceptable Use Agreement (printable version in PDF format)
and the supplemental AUA below:
When referencing this data (as required by the AUA), please use:The CAIDA UCSD Network Telescope Aggregated Flow Dataset - < dates used >,Also, please, report your publication to CAIDA.
UCSD Network Telescope Datasets
- Historical and Near-Real-Time Network Telescope Dataset
- Aggregated Traffic Data in FlowTuple format
- Daily RSDoS Attack Metadata
- Two Years of Daily RSDoS Attack Metadata (downloadable paper supplement)
- Three Days Of Conficker Dataset
- CAIDA UCSD Network Telescope Traffic Samples
- Witty Worm Dataset
- Code-Red Worms Dataset
- Patch Tuesday Dataset
- Two Days in November 2008 Dataset
- Telescope Educational Dataset
- Telescope Dataset on the Sipscan
- Telescope Darknet Scanners Dataset
For more information on the UCSD Network Telescope, see:
For more information on the CoralReef Software Suite, see:
For more information on the Corsaro Software Suite, see:
For a non-exhaustive list of Non-CAIDA publications using Network Telescope data, see: