Suduta: Script UAF detection using taint analysis

Abstract

Use-after-free (UAF) vulnerabilities are caused by the use of dangling pointers. Their exploitation inside script engine-hosting applications, e.g. web browsers, can even bypass state-of-the-art countermeasures. This work proposes SUDUTA (Script UAF Detection Using Taint Analysis), which aims at facilitating the diagnosis of UAF bugs during vulnerability analysis and improves an existent promising technique based on dynamic taint tracking. Firstly, precise taint analysis rules are presented in this work to clearly specify how SUDUTA manages the taint state. Moreover, it shifts its analysis to on-line, enabling instrumentation code to gain access to the program state of the application. Lastly, it handles the presence of custom memory allocators that are typically utilised in script-hosting applications. Results obtained using a benchmark dataset and vulnerable applications validate these three improvements.