Automated Binary Analysis on iOS
This is the implementation of our framework to automatically analyze iOS applications, as presented at WiSec 2018. See the paper by Feichtner, Missmann, and Spreitzer for more details.
In this repository you find our solution to:
- Decompile 64-bit ARM binaries to LLVM intermediate representation (IR),
- Perform static program slicing with pointer analysis on IR code, and
- Security rules to highlight improper usage of cryptographic APIs.
Note: This code is provided as-is. You are responsible for protecting yourself, your property and data, and others from any risks caused by this code. It may or may not detect vulnerabilities in your application/OS or device. It is intended only for educational purposes.
Setup
Important:: You might need macOS to get the whole thing running...
git clone https://github.com/IAIK/ios-analysis
cd ios-analysis
git submodule update --init --recursive
dagger
cd dagger
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE="Release" -DLLVM_TARGETS_TO_BUILD="AArch64;X86" ..
make -j4 llvm-dec
llvm-slicer
cd llvm-slicer
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE="Release" -DLLVM_TARGETS_TO_BUILD="AArch64;X86" -DLLVM_ENABLE_EH=YES -DLLVM_ENABLE_RTTI=ON ..
make -j4 opt
make -j4 llvm-slicer # requires that 'opt' is already built
Analysis Workflow
The basic steps are as follows:
- Get the .ipa file: E.g. using the tool Clutch or r2Clutch to decrypt the application you would like to inspect.
- Extract the ARM binary: Unzip an application package and extract the ARM binary.
extract.sh can be run in a folder with .ipa files. It extracts the individual app binaries and stores them in a new folder named
Bins
. - Start the analysis:
See run.sh for the steps needed to process a particular binary. The script first gets the ARMv8 64-bit file using
lipo
, decompiles it using dagger, applies some LLVM optimization passes, and finally feeds it to llvm-slicer in order to evaluate the rules specified in rules.json. The script takes the following arguments:
./run.sh {64-bit binary name} {optional: JSON rule file (default: $RULES/rules.json)} {optional: report output filename (default: $binary.html)}
dagger
llvm-dec should be called like this:
llvm-dec {64-bit binary input filename} -O1 -bc -o {IR Output}
- bc: sets the output format to bitcode. The default format is LLVM assembler code, but this increases the file size.
- O1: Optimizations to prepare the file for the usage in LLVMSlicer
llvm-slicer
Additionally to the optimizations already performed in llvm-dec
, the bitcode has to be prepared for LLVMSlicer using:
opt -instcombine -sroa -simplifycfg -constprop {input} -o {output}
The LLVMSlicer is then executed by calling:
llvm-slicer {IR Code} -binary {64-bit binary} -o /dev/null -r {report filename} -rules {Rules JSON file}