Pith. sign in

REVIEW

Blind De-anonymization Attacks using Social Networks

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1801.05534 v1 pith:6E7EZN4E submitted 2018-01-17 cs.SI

Blind De-anonymization Attacks using Social Networks

classification cs.SI
keywords de-anonymizationattacksdatainformationtechniquesattackpriorrequire
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

It is important to study the risks of publishing privacy-sensitive data. Even if sensitive identities (e.g., name, social security number) were removed and advanced data perturbation techniques were applied, several de-anonymization attacks have been proposed to re-identify individuals. However, existing attacks have some limitations: 1) they are limited in de-anonymization accuracy; 2) they require prior seed knowledge and suffer from the imprecision of such seed information. We propose a novel structure-based de-anonymization attack, which does not require the attacker to have prior information (e.g., seeds). Our attack is based on two key insights: using multi-hop neighborhood information, and optimizing the process of de-anonymization by exploiting enhanced machine learning techniques. The experimental results demonstrate that our method is robust to data perturbations and significantly outperforms the state-of-the-art de-anonymization techniques by up to $10\times$ improvement.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.