SFuzz

Implementation(Source code) of paper:

SFuzz: Slice-based Fuzzing for Real-Time Operating Systems. In Proceedings of the 29th ACM Conference on Computer and Communications Security (ACM CCS 2022).

Overview of SFuzz

Setup

To get the docker image directly, go to: https://drive.google.com/file/d/1gSw9QFLuntsqoQlI_KjzoI8bej8ZT-E7/view?usp=share_link

Static Analysis

Perform taint analysis on the specified firmware, slice and patch the program for the taint analysis results.

This part uses the firmware collection in the evaluation_set folder as input.

Instructions for running this tool

Download Ghidra (we use version 9.2.3).
Change MAXMEM=2G to MAXMEM=4G in the analyzeHeadless file in the support directory of Ghidra folder.
sudo apt install openjdk-11-jdk
Place each file in a specific directory according to run.sh or modify run.sh according to the file location.
Run run.sh for a single firmware or run run_all.py for the entire firmware set.

Directories

├── README.md
├── findbase         #  Tool to discover the base address for firmware loading
├── unstrip          #  Tool for recovering the symbols of a specific function in unsigned binary
├── unstrip_from_log #  Tools for recovering function symbols based on information in the log function
├── statistics_script#  Scripts for supporting statistics
├── evaluation_set   #  Unpacked firmware collection
├── findtrace_output #  Static analysis results for firmware in evaluation_set
├── findtrace.py     #  Ghidra script that performs taint analysis and outputs slices and patches
├── run.sh           #  Shell scripts for processing individual firmware
├── run_all.py       #  Python script for batch processing of multiple firmware
└── setbase.py       #  Ghidra script for setting the base address for firmware loading

Run

For processing individual firmware

Be sure to modify run.sh according to the actual file location before you run it

sudo ./run.sh firmware_path  arch  base_addr

example:

sudo ./run.sh evaluation_set/DIR-100/30_DIR100 MIPS:BE:32:default 0x80000100

For batch processing of multiple firmware

sudo python3 run_all.py

Output

Most of the files end with an underscore and a number, which represents the number of the call tree corresponding to the current file.

30_DIR100_result/
├── call_checksum_0  # Address of checksum function calls
├── .......
├── call_checksum_7
├── calltrace_0      # Results of call trace in call tree
├── ........
├── calltrace_7
├── cbranch_info_0   # Jumping information at the branch
├── ........
├── cbranch_info_7
├── dict_0           # String information on the call tree (used to add to the AFL dictionary)
├── ........
├── dict_7
├── exec_0           # Contextual information used for fuzzing (source address and sink address, etc.)
├── ........
├── exec_7
├── patch_0          # Patch results for function calls or branches
├── ........
├── patch_7
├── sink_buf_0       # Information about the sink function address and its corresponding buffer
├── ........
├── sink_buf_7
├── stack_retaddr_0  # Return address information on the stack
├── ........
├── stack_retaddr_7
├── summary          # Statistical Information
├── summary.json
├── xalloc_0         # Cross-reference information for the alloc function
├── ........
└── xalloc_7

Dynamic Analysis

Dynamic analysis includes fuzzing and concolic solving.

This project relies on the output of the static analysis as input.

Running environment configuration

apt-get install -y libc6-armel-cross gcc-arm-linux-gnueabi libc6-mipsel-cross gcc-mipsel-linux-gnu
apt-get install -y build-essential python3-dev automake cmake git flex bison libglib2.0-dev libpixman-1-dev python3-setuptools python python-setuptools
apt-get install -y lld llvm llvm-dev clang
apt-get install -y gcc-$(gcc --version|head -n1|sed 's/\..*//'|sed 's/.* //')-plugin-dev libstdc++-$(gcc --version|head -n1|sed 's/\..*//'|sed 's/.* //')-dev
apt-get install -y python3-pip 

git clone https://github.com/AFLplusplus/AFLplusplus.git
cd AFLplusplus
make all && make install

git clone https://github.com/Battelle/afl-unicorn
cd afl-unicorn/unicorn_mode

wget https://bootstrap.pypa.io/ez_setup.py -O - | python
sed -i '120,122d' ./build_unicorn_support.sh
wget https://github.com/unicorn-engine/unicorn/archive/refs/tags/1.0.3.zip
unzip 1.0.3.zip && rm 1.0.3.zip && mv unicorn-1.0.3 unicorn
./build_unicorn_support.sh

pip3 install --upgrade "pip<21.0.0"
pip3 install pwntools==4.8.0
pip3 install angr==9.2.6
pip3 install tqdm

There are also minor modifications to the dependent projects.

https://amusing-aluminum-be0.notion.site/d0de68137f884c1984d4ae7c762b566f

Also, AFL requires: if in a docker environment, execute with root privileges outside the docker environment, otherwise execute the following statements directly with root privileges.

echo core > /proc/sys/kernel/core_pattern

build fuzz loader

in ./dynamic_analysis dir (in ./uniFuzzGo dir in docker):

make or make "UFDBG=-DUF_DEBUG -g"

run