With the advent of Semantic Web and Resource Description Framework (RDF), the web is likely to witness an unprecedented wealth of knowledge, resulting from seamless integration of various data sources. Data integration is one of the key features of RDF, however, absence of secure means for managing sensitive RDF data may prevent sharing of critical data altogether or may cause serious damage. Towards this end we present a language for sanitizing RDF graphs, which comprises a set of sanitization operations that transform a graph by concealing the sensitive data. These operations are modeled into a new SPARQL query form known as SANITIZE, which can also be leveraged towards fine grained access control and building advanced anonymization features.