A deliberately insecure RDF-based Semantic Web application framework for teaching SPARQL/ SPARUL injection attacks and defense mechanisms

摘要

The Semantic Web uses the Resource Description Framework (RDF) and the Simple Protocol and Query/Update Languages (SPARQL/SPARUL) as standardized logical data representation and manipulation models allowing machines to directly interpret data on the Web. As Semantic Web applications grow increasingly popular, new and challenging security threats emerge. Semantic query languages owing to their flexible nature introduce new vulnerabilities if secure programming practices are not followed. This makes them prone to both existing attacks such as command injection as well as novel attacks, making it necessary for application developers to understand the security risks involved when developing and deploying semantic applications. In this research, we have analyzed and categorized the possible SPARQL/SPARUL injection attacks to which semantic applications are vulnerable. Moreover, we have developed a deliberately insecure RDF-based Semantic Web application, called SemWebGoat - inspired by the open source vulnerable web application, WebGoat - which offers a realistic teaching and learning environment for exploiting SPARQL/ SPARUL oriented injection vulnerabilities. With the aim of teaching both developers and web administrators the art of protecting their Semantic Web applications, we have implemented web application firewall (WAF) rules using the popular open-source firewall -ModSecurity - and extended some penetration testing tools to detect and mitigate SPARQL/ SPARUL injections. For the evaluation, we conducted a user study to determine the usability of SemWebGoat attack lessons as well as a detection rate and false alarm analysis of our proposed firewall rules based on OWASP top-ten attack dataset. The results of the user study conclude that web developers are not normally familiar with the injection vulnerabilities demonstrated. The positive test results of our ModSecurity rule set show that it a suitable defense mechanism for protecting vulnerable Semantic Web application against injection attacks.