Umeå universitets logga

We analyze known deserialization exploits targeting applications developed in the Java programming language. As previous research implies, fully comprehending this type of vulnerability is no easy task due to the complexity of exploitation, mostly relying on so-called gadget chains. Even considering known gadget chains, knowledge about their prerequisites is rather limited. In particular, the full range of external library versions, adding exploitable gadgets to the Java classpath was formerly only partially examined. We contribute an in-depth analysis of publicly available Java deserialization vulnerabilities. Specifically, we experimentally assess the prerequisites for exploitation, using 46 different gadget chains on 244 JDK and 5,455 Java dependency versions. Previous research only covered 19 of these gadget chains. Furthermore, we develop a command line tool, Gadgecy, for lightweight detection of whether a given Java project contains dependency combinations that enable gadget chains. Using this tool, we conduct an analysis of 2,211 projects from the Apache Distribution directory and 400 well-known Github repositories. The outcome reveals that (1) deserialization exploits apply to recent JDK and library versions, (2) these gadget chains are not being fully reported, and (3) are frequently present in popular Java projects (such as Apache Kafka or Hadoop).