Social-networking users unknowingly reveal certain kinds of personal information that malicious attackers could profit from to perpetrate significant privacy breaches. This paper quantitatively demonstrates how the simple act of tagging pictures on the social-networking site of Facebook could reveal private user attributes that are extremely sensitive. Our results suggest that photo tags can be used to help predicting some, but not all, of the analyzed attributes. We believe our analysis make users aware of significant breaches of their privacy and could inform the design of new privacy-preserving ways of tagging pictures on social-networking sites.