Optimized combined-clustering methods for finding replicated criminal websites

EURASIP Journal on Information Security

Table 1 Adjusted Rand index for different clusterings, varying the number of input attributes considered (best-performing clusterings italicized)

Scam websites	Dynamic cut height		Optimized cut height
	Test	Train	Test	Train
Fake escrow services
Sentences	0.107	0.289	0.982	0.924
DOM tags	0.678	0.648	0.979	0.919
File names	0.094	0.235	0.972	0.869
Images	0.068	0.206	0.325	0.314
S and D	0.942	0.584	0.982	0.925
S and F	0.120	0.245	0.980	0.895
S and I	0.072	0.257	0.962	0.564
D and F	0.558	0.561	0.979	0.892
D and I	0.652	0.614	0.599	0.385
F and I	0.100	0.224	0.518	0.510
S and D and F	0.913	0.561	0.980	0.895
S and D and I	0.883	0.536	0.971	0.673
S and F and I	0.100	0.214	0.975	0.892
D and F and I	0.642	0.536	0.831	0.772
S and D and F and I	0.941	0.536	0.971	0.683
High-yield investment programs
Sentences	0.713	0.650	0.738	0.867
DOM tags	0.381	0.399	0.512	0.580
File names	0.261	0.299	0.254	0.337
Images	0.289	0.354	0.434	0.471
S and D	0.393	0.369	0.600	0.671
S and F	0.291	0.310	0.266	0.344
S and I	0.290	0.362	0.437	0.471
D and F	0.309	0.358	0.314	0.326
D and I	0.302	0.340	0.456	0.510
F and I	0.296	0.289	0.397	0.336
S and D and F	0.333	0.362	0.319	0.326
S and D and I	0.319	0.350	0.459	0.510
S and F and I	0.303	0.289	0.398	0.336
D and F and I	0.320	0.337	0.404	0.405
S and D and F and I	0.320	0.337	0.404	0.405