摘要: In order to address privacy concerns, many social media websites allow users hide their personal profiles from the public. this work, we show how an adversary can exploit online network with a mixture of public and private user predict attributes users. We map problem relational classification propose practical models that use friendship group membership information (which is often not hidden) infer sensitive attributes. The key novel idea in addition links, groups be carriers significant information. on several well-known sites, easily accurately recover private-profile To best our knowledge, first work uses link-based group-based study implications networks mixed profiles.