Address Obfuscation: An Efficient Approach to Combat a Broad Range of Memory Error Exploits

Attacks which exploit memory programming errors (such as buffer overflows) are one of today's most serious security threats. These attacks require an attacker to have an in-depth understanding of the internal details of a victim program, including the locations of critical data and/or code. Program obfuscation is a general technique for securing programs by making it difficult for attackers to acquire such a detailed understanding. This paper develops a systematic study of a particular kind of obfuscation called address obfuscation that randomizes the location of victim program data and code. We discuss different implementation strategies to randomize the absolute locations of data and code, as well as relative distances between data locations. We then present our implementation that transforms object files and executables at link-time and load-time. It requires no changes to the OS kernel or compilers, and can be applied to individual applications without affecting the rest of the system. It can be implemented with low runtime overheads. Address obfuscation can reduce the probability of successful attacks to be as low as a small fraction of a percent for most memory-error related attacks. Moreover, the randomization ensures that an attack that succeeds against one victim will likely not succeed against another victim, or even for a second time against the same victim. Each failed attempt will typically crash the victim program, thereby making it easy to detect attack attempts. These aspects make it particularly effective against large-scale attacks such as Code Red, since each infection attempt requires significantly more resources, thereby slowing down the propagation rate of such attacks.