Skip to content
forked from harvey-lau/fgo

FGo: A Directed Grey-box Fuzzer with Probabilistic Exponential cut-the-loss Strategies

Notifications You must be signed in to change notification settings

joshuay2022/fgo

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

The ReadMe of fgo

0x01 Introduction

FGo is a probabilistic exponential cut-the-loss directed grey-box fuzzer based on AFLGo. FGo terminates unreachable test cases early with exponentially increasing probability. Compared to other directed grey-box fuzzers, FGo makes full use of the unreachable information contained in iCFG and doesn't generate any additional overhead caused by reachability analysis. Moreover, it is easy to generalize to all Program Under Test (PUT). This strategy based on probability is perfectly adapted to the randomness of fuzzing.

The usage of FGo is similar to AFLGo. The only difference is that you should replace

$AFLGO/scripts/gen_distance_fast.py $SUBJECT $TMP_DIR xmllint

with

$FGO/distance/distance_generator.py $SUBJECT $TMP_DIR xmllint

in the step which generates distance files.

Of course, all environment variables should be changed from AFLGO (or aflgo) to FGO (or fgo).

FGo add two argument options -p and -P to afl-fuzz, where -p represents the preparation time of FGo and -P represents the cut-the-loss probability $p$.

  • During the preparation time period, FGo doesn't conduct the cut-the-loss procedure since the directed fuzzer needs to analyze the feedback for the first time in a while in order for better performance.
  • The cut-the-loss probability $p$ has the limit $0 \lt p \lt 1.0$ and its precision is $0.01$.

Read its paper for more details.

0x02 How to Use

For example, the following commands fuzz min1 step by step.

You can find the related information of min1 like the project version in this paper.

1. Preparation

cd ~/obj-fgo
cp -rf ~/projects/libming .
mv libming min1
cd ~/obj-fgo/min1
git checkout 6f91d1a

export FGO=~/fgo
export SUBJECT=$PWD
mkdir obj-fgo
mkdir obj-fgo/temp
export TMP_DIR=$PWD/obj-fgo/temp

echo $'main.c:350\nmain.c:265\nblocktypes.c:145\nparser.c:3345\nparser.c:3302\nparser.c:3068' > $TMP_DIR/BBtargets.txt

2. First Compilation with BBtargets.txt

export CC=$FGO/afl-clang-fast
export CXX=$FGO/afl-clang-fast++
export COPY_CFLAGS=$CFLAGS
export COPY_CXXFLAGS=$CXXFLAGS
export ADDITIONAL="-targets=$TMP_DIR/BBtargets.txt -outdir=$TMP_DIR -flto -fuse-ld=gold -Wl,-plugin-opt=save-temps"
export CFLAGS="$COPY_CFLAGS $ADDITIONAL"
export CXXFLAGS="$COPY_CXXFLAGS $ADDITIONAL"
export LDFLAGS=-lpthread

./autogen.sh
cd ~/obj-fgo/min1/obj-fgo
../configure --disable-shared --disable-freetype
make clean
make

cat $TMP_DIR/BBnames.txt | rev | cut -d: -f2- | rev | sort | uniq > $TMP_DIR/BBnames2.txt && mv $TMP_DIR/BBnames2.txt $TMP_DIR/BBnames.txt
cat $TMP_DIR/BBcalls.txt | sort | uniq > $TMP_DIR/BBcalls2.txt && mv $TMP_DIR/BBcalls2.txt $TMP_DIR/BBcalls.txt

3. The Generation of distance.cfg.txt

$FGO/distance/distance_generator.py $SUBJECT/obj-fgo/util $TMP_DIR swftophp

4. Second Compilation with distance.cfg.txt

export CFLAGS="$COPY_CFLAGS -distance=$TMP_DIR/distance.cfg.txt"
export CXXFLAGS="$COPY_CXXFLAGS -distance=$TMP_DIR/distance.cfg.txt"

../configure --disable-shared --disable-freetype
make clean
make

5. Seed

mkdir in
wget -P in http://condor.depaul.edu/sjost/hci430/flash-examples/swf/bumble-bee1.swf

6. Fuzzing

  • -p <pre_time>: pre_time is an integer along with its time unit (second s, minute m, hour h)
  • -P <prob>: prob is an integer representing the cut-the-loss probability, which is scaled to 0~100
cd ~/obj-fgo/min1/obj-fgo
~/fgo/afl-fuzz -m none -z exp -c 120m -i in -o out -t 5000+ -p 30m -P 20 ./util/swftophp @@

About

FGo: A Directed Grey-box Fuzzer with Probabilistic Exponential cut-the-loss Strategies

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C 82.9%
  • Shell 6.8%
  • Python 3.5%
  • C++ 3.4%
  • Makefile 2.7%
  • HTML 0.7%