Allatori vs Obfuscator-LLVM

Allatori is a Java obfuscator that works on compiled class files to protect intellectual property and make reverse engineering more difficult. It does not require original Java source code. Its features include name, control-flow and debug-information obfuscation, string encryption, watermarking and incremental obfuscation, along with a utility for stack traces and a module intended to disrupt decompilers. It supports J2EE, J2SE, J2ME and Android targets. Build options include a command-line interface, Apache Ant, Maven, Android Studio and an Eclipse plugin. Written in Java, it can technically run wherever a Java 1.4-or-later virtual machine is supported; the maker says it has been tested on Windows, Mac OS X and UNIX. The maker reports typical application size reductions of about 10–40%, sometimes up to 60% with many third-party libraries. String encryption can add 1–2% to bytecode size, while flow obfuscation can slightly increase size and rarely reduce performance. Native methods and their containing classes are not renamed automatically. A single developer license costs $290 as a one-time purchase; the free plan is also listed.

Obfuscator-LLVM is a free, open-source LLVM fork for Linux that applies obfuscation while software is compiled. It is intended to make programs harder to reverse engineer or modify. Listed passes include instruction substitution, bogus control flow, and control-flow flattening, with function annotations also available. Its intermediate-representation approach covers languages supported by LLVM, including C, C++, Objective-C, Ada, and Fortran. Listed targets include x86, x86-64, PowerPC, ARM, Thumb, SPARC, Alpha, CellSPU, MIPS, MSP430, SystemZ, and XCore. Installation involves cloning the source repository, then building with CMake and make; binaries are placed in build/bin. The guide describes compiler use through Clang, autotools, and a custom Xcode plugin. The latest official non-development branch is llvm-4.0, based on LLVM 4.0.1. The project says obfuscated code can be slower and larger, and describes itself as an advanced prototype needing more testing before production use.