Geolocation of covert communication entity on the Internet for post-steganalysis

EURASIP Journal on Image and Video Processing

Table 3 Relationship between different training set sizes, different thresholds and the quantity of the entities that can be geolocated and geolocation error

Landmark sets	U^★	QCG/QCNG^●	MGE^◆
Training set size^§ > 100 (a total of 40 training sets and 41,231 landmarks)	0.9	9321/1365	2.80 km
	0.8	10109/577	4.62 km
	0.7	10549/137	6.16 km
Training set size > 300 (a total of 14 training sets and 37,134 landmarks)	0.9	8203/2483	2.26 km
	0.8	8748/1938	2.97 km
	0.7	9108/1578	4.05 km
Training set size > 500 (a total of 10 training sets and 32,241 landmarks)	0.9	7296/3390	2.18 km
	0.8	8052/2634	2.51 km
	0.7	8724/1962	2.96 km

^§“Training set size > N” represents a landmark set composed of all training sets with a landmark quantity greater than N in the training set
^★U is short for “geolocation threshold”
^●QCG is short for “quantity of the entities that can be geolocated”
QCNG is short for “quantity of the entities that cannot be geolocated”
^◆MGE is short for “median geolocation error of the entities that can be geolocated”