Understanding structure-based social network de-anonymization techniques via empirical analysis

EURASIP Journal on Wireless Communications and Networking

Table 1 Comparison of representative de-anonymization methods

DA approach	Vulnerable Anonymization Method			Performance				Category
	Edge-edit	k-Anony.	U-Split	Para.-free	Scalable	Practical	Robust	I	II	III	IV
Backstrom et al. [11]	×	×	×	√	×	∙	×	√
N-DA. [4]	√	∙	∙	×	√	√	√	√
Srivatsa et al.-DV [5]	√	∙	∙	√	∙	∙	√		√
Nilizadeh et al. [19]	∙	∙	∙	×	∙	∙	∙	√
Ji-DA [37]	√	∙	∙	×	√	√	√		√
Ji-ADA [37]	√	∙	∙	×	√	√	√		√
Ji-ODA [38]	√	∙	∙	×	√	√	√			√
Chen et al. [42]	√	∙	∙	√	√	√	√				√
Ji-DeSAG [28]	√	∙	∙	×	√	√	√				√

As for vulnerability of anonymization methods, √ denotes that the attack can succeed under corresponding anonymization methods, × denotes the attack’s failure, and ∙ denotes the attack can conditionally succeed. As for performance, √ represents that the attack is capable of corresponding character, × represents the attack’s incapability, and ∙ represents the attack is conditionally capable. Category I: structural user-based attacks with seed. Category II: structural set-based attacks with seed. Category III: structural attacks without seed. Category IV: structure-attribute-based attacks
k-Anony. k-anonymity, U-Split union-split, Para. parameters