| Coarse-grained Inference of BGP Community Intent |
| Welcome |
| Inferences |
|
We provide our mapping of coarse-grained categories (action and informational) to communities observed in public BGP data.
Download infered categories: 2023050x.communityinference.txt.bz2 (214K) The mappings are created using our inference method and 1 week (May 1-7, 2023) of RIBs and updated provided by RouteViews and RIPE RIS.
$ zcat 2023050x.communities.txt.gz | head -n8
1:0,action
1:1,action
1:2,action
1:3,action
1:9,action
1:10,action
1:11,action
1:12,action
$ zcat 2023050x.communities.txt.gz | cut -d"," -f2 | sort | uniq -c
24376 action
54104 info
|
| Reproducability |
|
In order to reproduce the inferences above, you will need the following files:
communityinference.py (11K) 2023050x.onoff.txt.bz2 (829K) 2023050x.asns.txt.bz2 (182K) 2023050x.labels.txt.bz2 (16K) Download those files and use them in python3 as followed (or download the jupyter notebook: communityinference.ipynb):
mingap = 140
onoffthresh = 160
exec(open("communityinference.py").read())
mymodel = CommunityInference()
mymodel.load_feats('2023050x.onoff.txt.bz2')
mymodel.load_asns('2023050x.asns.txt.bz2')
mymodel.augment_feats('2023050x.labels.txt.bz2')
mymodel.augment_clusters(mingap)
df_comm, df_clust = mymodel.infer_default(onoffthresh)
df_comm[['comm','pred']].to_csv('2023050x.communityinference.txt.bz2', index=False, header=False)
|