On Detecting Code Reuse Attacks

摘要

Today, a code reuse technique is often used when exploiting software vulnerabilities, such as a buffer overflow. These attacks bypass the protection against execution of code in the stack, which is implemented on the hardware and software levels in modern information systems. The attacks are based on finding suitable sections of executable code-gadgets-in the vulnerable program and linking these gadgets into chains. The article proposes a method to protect applications against code reuse attacks. The method is based on detecting properties that distinguish between chains of gadgets and typical chains of legitimate program basic blocks. The appearance of an atypical chain of basic blocks during program execution may indicate the execution of a malicious code. One of the properties of a chain of gadgets is that at the end of the chain a special processor instruction used to call a function of the operating system is executed. Experiments are carried out for the x86/64 Linux operating system which show the importance of this property for detecting malicious code execution. An algorithm for identifying atypical chains is developed which makes it possible to detect all currently known code reuse techniques.