Fuzzing with AFL++

The following describes how to fuzz with a target if source code is available. If you have a binary-only target, go to /docs/fuzzing_binary-only_targets/.

Fuzzing source code is a three-step process:

Compile the target with a special compiler that prepares the target to be fuzzed efficiently. This step is called “instrumenting a target”.
Prepare the fuzzing by selecting and optimizing the input corpus for the target.
Perform the fuzzing of the target by randomly mutating input and assessing if that input was processed on a new path in the target binary.

0. Common sense risks

Please keep in mind that, similarly to many other computationally-intensive tasks, fuzzing may put a strain on your hardware and on the OS. In particular:

Your CPU will run hot and will need adequate cooling. In most cases, if cooling is insufficient or stops working properly, CPU speeds will be automatically throttled. That said, especially when fuzzing on less suitable hardware (laptops, smartphones, etc.), it’s not entirely impossible for something to blow up.
Targeted programs may end up erratically grabbing gigabytes of memory or filling up disk space with junk files. AFL++ tries to enforce basic memory limits, but can’t prevent each and every possible mishap. The bottom line is that you shouldn’t be fuzzing on systems where the prospect of data loss is not an acceptable risk.
Fuzzing involves billions of reads and writes to the filesystem. On modern systems, this will be usually heavily cached, resulting in fairly modest “physical” I/O - but there are many factors that may alter this equation. It is your responsibility to monitor for potential trouble; with very heavy I/O, the lifespan of many HDDs and SSDs may be reduced.

A good way to monitor disk I/O on Linux is the iostat command:
```
$ iostat -d 3 -x -k [...optional disk ID...]
```
Using the AFL_TMPDIR environment variable and a RAM-disk, you can have the heavy writing done in RAM to prevent the aforementioned wear and tear. For example, the following line will run a Docker container with all this preset:
```
# docker run -ti --mount type=tmpfs,destination=/ramdisk -e AFL_TMPDIR=/ramdisk aflplusplus/aflplusplus
```

1. Instrumenting the target

a) Selecting the best AFL++ compiler for instrumenting the target

AFL++ comes with a central compiler afl-cc that incorporates various different kinds of compiler targets and instrumentation options. The following evaluation flow will help you to select the best possible.

It is highly recommended to have the newest llvm version possible installed, anything below 9 is not recommended.

+--------------------------------+
| clang/clang++ 11+ is available | --> use LTO mode (afl-clang-lto/afl-clang-lto++)
+--------------------------------+     see [https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.lto.md](https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.lto.md)
    |
    | if not, or if the target fails with LTO afl-clang-lto/++
    |
    v
+---------------------------------+
| clang/clang++ 3.8+ is available | --> use LLVM mode (afl-clang-fast/afl-clang-fast++)
+---------------------------------+     see [https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.llvm.md](https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.llvm.md)
    |
    | if not, or if the target fails with LLVM afl-clang-fast/++
    |
    v
 +--------------------------------+
 | gcc 5+ is available            | -> use GCC_PLUGIN mode (afl-gcc-fast/afl-g++-fast)
 +--------------------------------+    see [https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.gcc_plugin.md](https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.gcc_plugin.md) and
                                       [https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.instrument_list.md](https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/README.instrument_list.md)
    |
    | if not, or if you do not have a gcc with plugin support
    |
    v
   use GCC mode (afl-gcc/afl-g++) (or afl-clang/afl-clang++ for clang)

Clickable README links for the chosen compiler:

LTO mode - afl-clang-lto
LLVM mode - afl-clang-fast
GCC_PLUGIN mode - afl-gcc-fast
GCC/CLANG modes (afl-gcc/afl-clang) have no README as they have no own features

You can select the mode for the afl-cc compiler by one of the following methods:

Using a symlink to afl-cc: afl-gcc, afl-g++, afl-clang, afl-clang++, afl-clang-fast, afl-clang-fast++, afl-clang-lto, afl-clang-lto++, afl-gcc-fast, afl-g++-fast (recommended!).
Using the environment variable AFL_CC_COMPILER with MODE.
Passing –afl-MODE command line options to the compiler via CFLAGS/CXXFLAGS/CPPFLAGS.

MODE can be one of the following:

LTO (afl-clang-lto*)
LLVM (afl-clang-fast*)
GCC_PLUGIN (afl-g*-fast) or GCC (afl-gcc/afl-g++)
CLANG(afl-clang/afl-clang++)

Because no AFL++ specific command-line options are accepted (beside the –afl-MODE command), the compile-time tools make fairly broad use of environment variables, which can be listed with afl-cc -hh or looked up in /docs/env_variables/.

b) Selecting instrumentation options

If you instrument with LTO mode (afl-clang-fast/afl-clang-lto), the following options are available:

Splitting integer, string, float, and switch comparisons so AFL++ can easier solve these. This is an important option if you do not have a very good and large input corpus. This technique is called laf-intel or COMPCOV. To use this, set the following environment variable before compiling the target: export AFL_LLVM_LAF_ALL=1. You can read more about this in instrumentation/README.laf-intel.md.
A different technique (and usually a better one than laf-intel) is to instrument the target so that any compare values in the target are sent to AFL++ which then tries to put these values into the fuzzing data at different locations. This technique is very fast and good - if the target does not transform input data before comparison. Therefore, this technique is called input to state or redqueen. If you want to use this technique, then you have to compile the target twice, once specifically with/for this mode by setting AFL_LLVM_CMPLOG=1, and pass this binary to afl-fuzz via the -c parameter. Note that you can compile also just a cmplog binary and use that for both, however, there will be a performance penalty. You can read more about this in instrumentation/README.cmplog.md.

If you use LTO, LLVM, or GCC_PLUGIN mode (afl-clang-fast/afl-clang-lto/afl-gcc-fast), you have the option to selectively instrument parts of the target that you are interested in. For afl-clang-fast, you have to use an llvm version newer than 10.0.0 or a mode other than DEFAULT/PCGUARD.

This step can be done either by explicitly including parts to be instrumented or by explicitly excluding parts from instrumentation.

To instrument only specified parts, create a file (e.g., allowlist.txt) with all the filenames and/or functions of the source code that should be instrumented and then:
1. Just put one filename or function (prefixing with fun: ) per line (no directory information necessary for filenames) in the file allowlist.txt.
  
  Example:
```
foo.cpp        # will match foo/foo.cpp, bar/foo.cpp, barfoo.cpp etc.
fun: foo_func  # will match the function foo_func
```
2. Set export AFL_LLVM_ALLOWLIST=allowlist.txt to enable selective positive instrumentation.
Similarly to exclude specified parts from instrumentation, create a file (e.g., denylist.txt) with all the filenames of the source code that should be skipped during instrumentation and then:
1. Same as above. Just put one filename or function per line in the file denylist.txt.
2. Set export AFL_LLVM_DENYLIST=denylist.txt to enable selective negative instrumentation.

NOTE: During optimization functions might be inlined and then would not match the list! See instrumentation/README.instrument_list.md.

There are many more options and modes available, however, these are most of the time less effective. See:

AFL++ performs “never zero” counting in its bitmap. You can read more about this here:

instrumentation/README.llvm.md#8-neverzero-counters

c) Selecting sanitizers

It is possible to use sanitizers when instrumenting targets for fuzzing, which allows you to find bugs that would not necessarily result in a crash.

Note that sanitizers have a huge impact on CPU (= less executions per second) and RAM usage. Also, you should only run one afl-fuzz instance per sanitizer type. This is enough because e.g. a use-after-free bug will be picked up by ASAN (address sanitizer) anyway after syncing test cases from other fuzzing instances, so running more than one address sanitized target would be a waste.

The following sanitizers have built-in support in AFL++:

ASAN = Address SANitizer, finds memory corruption vulnerabilities like use-after-free, NULL pointer dereference, buffer overruns, etc. Enabled with export AFL_USE_ASAN=1 before compiling.
MSAN = Memory SANitizer, finds read accesses to uninitialized memory, e.g., a local variable that is defined and read before it is even set. Enabled with export AFL_USE_MSAN=1 before compiling.
UBSAN = Undefined Behavior SANitizer, finds instances where - by the C and C++ standards - undefined behavior happens, e.g., adding two signed integers where the result is larger than what a signed integer can hold. Enabled with export AFL_USE_UBSAN=1 before compiling.
CFISAN = Control Flow Integrity SANitizer, finds instances where the control flow is found to be illegal. Originally this was rather to prevent return oriented programming (ROP) exploit chains from functioning. In fuzzing, this is mostly reduced to detecting type confusion vulnerabilities - which is, however, one of the most important and dangerous C++ memory corruption classes! Enabled with export AFL_USE_CFISAN=1 before compiling.
TSAN = Thread SANitizer, finds thread race conditions. Enabled with export AFL_USE_TSAN=1 before compiling.
LSAN = Leak SANitizer, finds memory leaks in a program. This is not really a security issue, but for developers this can be very valuable. Note that unlike the other sanitizers above this needs __AFL_LEAK_CHECK(); added to all areas of the target source code where you find a leak check necessary! Enabled with export AFL_USE_LSAN=1 before compiling. To ignore the memory-leaking check for certain allocations, __AFL_LSAN_OFF(); can be used before memory is allocated, and __AFL_LSAN_ON(); afterwards. Memory allocated between these two macros will not be checked for memory leaks.

It is possible to further modify the behavior of the sanitizers at run-time by setting ASAN_OPTIONS=..., LSAN_OPTIONS etc. - the available parameters can be looked up in the sanitizer documentation of llvm/clang. afl-fuzz, however, requires some specific parameters important for fuzzing to be set. If you want to set your own, it might bail and report what it is missing.

Note that some sanitizers cannot be used together, e.g., ASAN and MSAN, and others often cannot work together because of target weirdness, e.g., ASAN and CFISAN. You might need to experiment which sanitizers you can combine in a target (which means more instances can be run without a sanitized target, which is more effective).

d) Modifying the target

If the target has features that make fuzzing more difficult, e.g., checksums, HMAC, etc., then modify the source code so that checks for these values are removed. This can even be done safely for source code used in operational products by eliminating these checks within these AFL++ specific blocks:

#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
  // say that the checksum or HMAC was fine - or whatever is required
  // to eliminate the need for the fuzzer to guess the right checksum
  return 0;
#endif

All AFL++ compilers will set this preprocessor definition automatically.

e) Instrumenting the target

In this step, the target source code is compiled so that it can be fuzzed.

Basically, you have to tell the target build system that the selected AFL++ compiler is used. Also - if possible - you should always configure the build system in such way that the target is compiled statically and not dynamically. How to do this is described below.

The #1 rule when instrumenting a target is: avoid instrumenting shared libraries at all cost. You would need to set LD_LIBRARY_PATH to point to these, you could accidentally type “make install” and install them system wide - so don’t. Really don’t. Always compile libraries you want to have instrumented as static and link these to the target program!

Then build the target. (Usually with make.)

NOTES

Sometimes configure and build systems are fickle and do not like stderr output (and think this means a test failure) - which is something AFL++ likes to do to show statistics. It is recommended to disable AFL++ instrumentation reporting via export AFL_QUIET=1.
Sometimes configure and build systems error on warnings - these should be disabled (e.g., --disable-werror for some configure scripts).
In case the configure/build system complains about AFL++’s compiler and aborts, then set export AFL_NOOPT=1 which will then just behave like the real compiler and run the configure step separately. For building the target afterwards this option has to be unset again!

configure

For configure build systems, this is usually done by:

CC=afl-clang-fast CXX=afl-clang-fast++ ./configure --disable-shared

Note that if you are using the (better) afl-clang-lto compiler, you also have to set AR to llvm-ar[-VERSION] and RANLIB to llvm-ranlib[-VERSION] - as is described in instrumentation/README.lto.md.

CMake

For CMake build systems, this is usually done by:

mkdir build; cd build; cmake -DCMAKE_C_COMPILER=afl-cc -DCMAKE_CXX_COMPILER=afl-c++ ..

Note that if you are using the (better) afl-clang-lto compiler you also have to set AR to llvm-ar[-VERSION] and RANLIB to llvm-ranlib[-VERSION] - as is described in instrumentation/README.lto.md.

Meson Build System

For the Meson Build System, you have to set the AFL++ compiler with the very first command!

CC=afl-cc CXX=afl-c++ meson

Other build systems or if configure/cmake didn’t work

Sometimes cmake and configure do not pick up the AFL++ compiler or the RANLIB/AR that is needed - because this was just not foreseen by the developer of the target. Or they have non-standard options. Figure out if there is a non-standard way to set this, otherwise set up the build normally and edit the generated build environment afterwards manually to point it to the right compiler (and/or RANLIB and AR).

In complex, weird, alien build systems you can try this neat project: https://github.com/fuzzah/exeptor

Linker scripts

If the project uses linker scripts to hide the symbols exported by the binary, then you may see errors such as:

undefined symbol: __afl_area_ptr

The solution is to modify the linker script to add:

{
  global:
    __afl_*;
}

f) Better instrumentation

If you just fuzz a target program as-is, you are wasting a great opportunity for much more fuzzing speed.

This variant requires the usage of afl-clang-lto, afl-clang-fast or afl-gcc-fast.

It is the so-called persistent mode, which is much, much faster but requires that you code a source file that is specifically calling the target functions that you want to fuzz, plus a few specific AFL++ functions around it. See instrumentation/README.persistent_mode.md for details.

Basically, if you do not fuzz a target in persistent mode, then you are just doing it for a hobby and not professionally :-).

g) libfuzzer fuzzer harnesses with LLVMFuzzerTestOneInput()

libfuzzer LLVMFuzzerTestOneInput() harnesses are the defacto standard for fuzzing, and they can be used with AFL++ (and honggfuzz) as well!

Compiling them is as simple as:

afl-clang-fast++ -fsanitize=fuzzer -o harness harness.cpp targetlib.a

You can even use advanced libfuzzer features like FuzzedDataProvider, LLVMFuzzerInitialize() etc. and they will work!

The generated binary is fuzzed with afl-fuzz like any other fuzz target.

Bonus: the target is already optimized for fuzzing due to persistent mode and shared-memory test cases and hence gives you the fastest speed possible.

For more information, see utils/aflpp_driver/README.md.

2. Preparing the fuzzing campaign

As you fuzz the target with mutated input, having as diverse inputs for the target as possible improves the efficiency a lot.

a) Collecting inputs

To operate correctly, the fuzzer requires one or more starting files that contain a good example of the input data normally expected by the targeted application.

Try to gather valid inputs for the target from wherever you can. E.g., if it is the PNG picture format, try to find as many PNG files as possible, e.g., from reported bugs, test suites, random downloads from the internet, unit test case data - from all kind of PNG software.

If the input format is not known, you can also modify a target program to write normal data it receives and processes to a file and use these.

You can find many good examples of starting files in the testcases/ subdirectory that comes with this tool.

b) Making the input corpus unique

Use the AFL++ tool afl-cmin to remove inputs from the corpus that do not produce a new path/coverage in the target:

Put all files from step a into one directory, e.g., INPUTS.
Run afl-cmin:
- If the target program is to be called by fuzzing as bin/target INPUTFILE, replace the INPUTFILE argument that the target program would read from with @@:
```
afl-cmin -i INPUTS -o INPUTS_UNIQUE -- bin/target -someopt @@
```
- If the target reads from stdin (standard input) instead, just omit the @@ as this is the default:
```
afl-cmin -i INPUTS -o INPUTS_UNIQUE -- bin/target -someopt
```

This step is highly recommended, because afterwards the testcase corpus is not bloated with duplicates anymore, which would slow down the fuzzing progress!

c) Minimizing all corpus files

The shorter the input files that still traverse the same path within the target, the better the fuzzing will be. This minimization is done with afl-tmin, however, it is a long process as this has to be done for every file:

mkdir input
cd INPUTS_UNIQUE
for i in *; do
  afl-tmin -i "$i" -o "../input/$i" -- bin/target -someopt @@
done

This step can also be parallelized, e.g., with parallel.

Note that this step is rather optional though.

Done!

The INPUTS_UNIQUE/ directory from step b - or even better the directory input/ if you minimized the corpus in step c - is the resulting input corpus directory to be used in fuzzing! :-)

3. Fuzzing the target

In this final step, fuzz the target. There are not that many important options to run the target - unless you want to use many CPU cores/threads for the fuzzing, which will make the fuzzing much more useful.

If you just use one instance for fuzzing, then you are fuzzing just for fun and not seriously :-)

a) Running afl-fuzz

Before you do even a test run of afl-fuzz, execute sudo afl-system-config (on the host if you execute afl-fuzz in a Docker container). This reconfigures the system for optimal speed - which afl-fuzz checks and bails otherwise. Set export AFL_SKIP_CPUFREQ=1 for afl-fuzz to skip this check if you cannot run afl-system-config with root privileges on the host for whatever reason.

Note:

There is also sudo afl-persistent-config which sets additional permanent boot options for a much better fuzzing performance.
Both scripts improve your fuzzing performance but also decrease your system protection against attacks! So set strong firewall rules and only expose SSH as a network service if you use these (which is highly recommended).

If you have an input corpus from step 2, then specify this directory with the -i option. Otherwise, create a new directory and create a file with any content as test data in there.

If you do not want anything special, the defaults are already usually best, hence all you need is to specify the seed input directory with the result of step 2a) Collecting inputs:

afl-fuzz -i input -o output -- bin/target -someopt @@

Note that the directory specified with -o will be created if it does not exist.

It can be valuable to run afl-fuzz in a screen or tmux shell so you can log off, or afl-fuzz is not aborted if you are running it in a remote ssh session where the connection fails in between. Only do that though once you have verified that your fuzzing setup works! Run it like screen -dmS afl-main -- afl-fuzz -M main-$HOSTNAME -i ... and it will start away in a screen session. To enter this session, type screen -r afl-main. You see - it makes sense to name the screen session same as the afl-fuzz -M/-S naming :-) For more information on screen or tmux, check their documentation.

If you need to stop and re-start the fuzzing, use the same command line options (or even change them by selecting a different power schedule or another mutation mode!) and switch the input directory with a dash (-):

afl-fuzz -i - -o output -- bin/target -someopt @@

Adding a dictionary is helpful. You have the following options:

See the directory dictionaries/, if something is already included for your data format, and tell afl-fuzz to load that dictionary by adding -x dictionaries/FORMAT.dict.
With afl-clang-lto, you have an autodictionary generation for which you need to do nothing except to use afl-clang-lto as the compiler.
With afl-clang-fast, you can set AFL_LLVM_DICT2FILE=/full/path/to/new/file.dic to automatically generate a dictionary during target compilation. Adding AFL_LLVM_DICT2FILE_NO_MAIN=1 to not parse main (usually command line parameter parsing) is often a good idea too.
You also have the option to generate a dictionary yourself during an independent run of the target, see utils/libtokencap/README.md.
Finally, you can also write a dictionary file manually, of course.

afl-fuzz has a variety of options that help to workaround target quirks like very specific locations for the input file (-f), performing deterministic fuzzing (-D) and many more. Check out afl-fuzz -h.

We highly recommend that you set a memory limit for running the target with -m which defines the maximum memory in MB. This prevents a potential out-of-memory problem for your system plus helps you detect missing malloc() failure handling in the target. Play around with various -m values until you find one that safely works for all your input seeds (if you have good ones and then double or quadruple that).

By default, afl-fuzz never stops fuzzing. To terminate AFL++, press Control-C or send a signal SIGINT. You can limit the number of executions or approximate runtime in seconds with options also.

When you start afl-fuzz, you will see a user interface that shows what the status is:

All labels are explained in /docs/afl-fuzz_approach/#understanding-the-status-screen.

b) Keeping memory use and timeouts in check

Memory limits are not enforced by afl-fuzz by default and the system may run out of memory. You can decrease the memory with the -m option, the value is in MB. If this is too small for the target, you can usually see this by afl-fuzz bailing with the message that it could not connect to the forkserver.

Consider setting low values for -m and -t.

For programs that are nominally very fast, but get sluggish for some inputs, you can also try setting -t values that are more punishing than what afl-fuzz dares to use on its own. On fast and idle machines, going down to -t 5 may be a viable plan.

The -m parameter is worth looking at, too. Some programs can end up spending a fair amount of time allocating and initializing megabytes of memory when presented with pathological inputs. Low -m values can make them give up sooner and not waste CPU time.

c) Using multiple cores

If you want to seriously fuzz, then use as many cores/threads as possible to fuzz your target.

On the same machine - due to the design of how AFL++ works - there is a maximum number of CPU cores/threads that are useful, use more and the overall performance degrades instead. This value depends on the target, and the limit is between 32 and 64 cores per machine.

If you have the RAM, it is highly recommended run the instances with a caching of the test cases. Depending on the average test case size (and those found during fuzzing) and their number, a value between 50-500MB is recommended. You can set the cache size (in MB) by setting the environment variable AFL_TESTCACHE_SIZE.

There should be one main fuzzer (-M main-$HOSTNAME option - set also AFL_FINAL_SYNC=1) and as many secondary fuzzers (e.g., -S variant1) as you have cores that you use. Every -M/-S entry needs a unique name (that can be whatever), however, the same -o output directory location has to be used for all instances.

For every secondary fuzzer there should be a variation, e.g.:

one should fuzz the target that was compiled with sanitizers activated (export AFL_USE_ASAN=1 ; export AFL_USE_UBSAN=1 ; export AFL_USE_CFISAN=1)
one or two should fuzz the target with CMPLOG/redqueen (see above), at least one cmplog instance should follow transformations (-l 2AT)
one to three fuzzers should fuzz a target compiled with laf-intel/COMPCOV (see above). Important note: If you run more than one laf-intel/COMPCOV fuzzer and you want them to share their intermediate results, the main fuzzer (-M) must be one of them (although this is not really recommended).

The other secondaries should be run like this:

10% with the MOpt mutator enabled: -L 0
10% should use the old queue cycling with -Z
50-70% should run with AFL_DISABLE_TRIM
40% should run with -P explore and 20% with -P exploit
If you use -a then set 30% of the instances to not use -a; if you did not set -a (why??), then set 30% to -a ascii and 30% to -a binary.
run each with a different power schedule, recommended are: fast (default), explore, coe, lin, quad, exploit, and rare which you can set with the -p option, e.g., -p explore. See the FAQ for details.

It can be useful to set AFL_IGNORE_SEED_PROBLEMS=1 to skip over seeds that crash or timeout during startup.

Also, it is recommended to set export AFL_IMPORT_FIRST=1 to load test cases from other fuzzers in the campaign first. But note that can slow down the start of the first fuzz by quite a lot of you have many fuzzers and/or many seeds.

If you have a large corpus, a corpus from a previous run or are fuzzing in a CI, then also set export AFL_CMPLOG_ONLY_NEW=1 and export AFL_FAST_CAL=1. If the queue in the CI is huge and/or the execution time is slow then you can also add AFL_NO_STARTUP_CALIBRATION=1 to skip the initial queue calibration phase and start fuzzing at once - but only do this if the calibration phase would be too long for your fuzz run time.

You can also use different fuzzers. If you are using AFL spinoffs or AFL conforming fuzzers, then just use the same -o directory and give it a unique -S name. Examples are:

A long list can be found at https://github.com/Microsvuln/Awesome-AFL.

However, you can also sync AFL++ with honggfuzz, libfuzzer with -entropic=1, etc. Just show the main fuzzer (-M) with the -F option where the queue/work directory of a different fuzzer is, e.g., -F /src/target/honggfuzz. Using honggfuzz (with -n 1 or -n 2) and libfuzzer in parallel is highly recommended!

d) Using multiple machines for fuzzing

Maybe you have more than one machine you want to fuzz the same target on. Start the afl-fuzz (and perhaps libfuzzer, honggfuzz, …) orchestra as you like, just ensure that your have one and only one -M instance per server, and that its name is unique, hence the recommendation for -M main-$HOSTNAME.

Now there are three strategies on how you can sync between the servers:

never: sounds weird, but this makes every server an island and has the chance that each follow different paths into the target. You can make this even more interesting by even giving different seeds to each server.
regularly (~4h): this ensures that all fuzzing campaigns on the servers “see” the same thing. It is like fuzzing on a huge server.
in intervals of 1/10th of the overall expected runtime of the fuzzing you sync. This tries a bit to combine both. Have some individuality of the paths each campaign on a server explores, on the other hand if one gets stuck where another found progress this is handed over making it unstuck.

The syncing process itself is very simple. As the -M main-$HOSTNAME instance syncs to all -S secondaries as well as to other fuzzers, you have to copy only this directory to the other machines.

Let’s say all servers have the -o out directory in /target/foo/out, and you created a file servers.txt which contains the hostnames of all participating servers, plus you have an ssh key deployed to all of them, then run:

for FROM in `cat servers.txt`; do
  for TO in `cat servers.txt`; do
    rsync -rlpogtz --rsh=ssh $FROM:/target/foo/out/main-$FROM $TO:target/foo/out/
  done
done

You can run this manually, per cron job - as you need it. There is a more complex and configurable script in utils/distributed_fuzzing.

e) The status of the fuzz campaign

AFL++ comes with the afl-whatsup script to show the status of the fuzzing campaign.

Just supply the directory that afl-fuzz is given with the -o option and you will see a detailed status of every fuzzer in that campaign plus a summary.

To have only the summary, use the -s switch, e.g., afl-whatsup -s out/.

If you have multiple servers, then use the command after a sync or you have to execute this script per server.

Another tool to inspect the current state and history of a specific instance is afl-plot, which generates an index.html file and graphs that show how the fuzzing instance is performing. The syntax is afl-plot instance_dir web_dir, e.g., afl-plot out/default /srv/www/htdocs/plot.

f) Stopping fuzzing, restarting fuzzing, adding new seeds

To stop an afl-fuzz run, press Control-C.

To restart an afl-fuzz run, just reuse the same command line but replace the -i directory with -i - or set AFL_AUTORESUME=1.

If you want to add new seeds to a fuzzing campaign, you can run a temporary fuzzing instance, e.g., when your main fuzzer is using -o out and the new seeds are in newseeds/ directory:

AFL_BENCH_JUST_ONE=1 AFL_FAST_CAL=1 afl-fuzz -i newseeds -o out -S newseeds -- ./target

g) Checking the coverage of the fuzzing

The corpus count value is a bad indicator for checking how good the coverage is.

A better indicator - if you use default llvm instrumentation with at least version 9 - is to use afl-showmap with the collect coverage option -C on the output directory:

$ afl-showmap -C -i out -o /dev/null -- ./target -params @@
...
[*] Using SHARED MEMORY FUZZING feature.
[*] Target map size: 9960
[+] Processed 7849 input files.
[+] Captured 4331 tuples (highest value 255, total values 67130596) in '/dev/nul
l'.
[+] A coverage of 4331 edges were achieved out of 9960 existing (43.48%) with 7849 input files.

It is even better to check out the exact lines of code that have been reached - and which have not been found so far.

An “easy” helper script for this is https://github.com/vanhauser-thc/afl-cov, just follow the README of that separate project.

If you see that an important area or a feature has not been covered so far, then try to find an input that is able to reach that and start a new secondary in that fuzzing campaign with that seed as input, let it run for a few minutes, then terminate it. The main node will pick it up and make it available to the other secondary nodes over time. Set export AFL_NO_AFFINITY=1 or export AFL_TRY_AFFINITY=1 if you have no free core.

Note that in nearly all cases you can never reach full coverage. A lot of functionality is usually dependent on exclusive options that would need individual fuzzing campaigns each with one of these options set. E.g., if you fuzz a library to convert image formats and your target is the png to tiff API, then you will not touch any of the other library APIs and features.

h) How long to fuzz a target?

This is a difficult question. Basically, if no new path is found for a long time (e.g., for a day or a week), then you can expect that your fuzzing won’t be fruitful anymore. However, often this just means that you should switch out secondaries for others, e.g., custom mutator modules, sync to very different fuzzers, etc.

Keep the queue/ directory (for future fuzzings of the same or similar targets) and use them to seed other good fuzzers like libfuzzer with the -entropic switch or honggfuzz.

i) Improve the speed!

Use persistent mode (x2-x20 speed increase).
If you do not use shmem persistent mode, use AFL_TMPDIR to point the input file on a tempfs location, see /docs/env_variables/.
Linux: Improve kernel performance: modify /etc/default/grub, set GRUB_CMDLINE_LINUX_DEFAULT="ibpb=off ibrs=off kpti=off l1tf=off mds=off mitigations=off no_stf_barrier noibpb noibrs nopcid nopti nospec_store_bypass_disable nospectre_v1 nospectre_v2 pcid=off pti=off spec_store_bypass_disable=off spectre_v2=off stf_barrier=off"; then update-grub and reboot (warning: makes the system more insecure) - you can also just run sudo afl-persistent-config.
Linux: Running on an ext2 filesystem with noatime mount option will be a bit faster than on any other journaling filesystem.
Use your cores! See 3c) Using multiple cores.
Run sudo afl-system-config before starting the first afl-fuzz instance after a reboot.

j) Going beyond crashes

Fuzzing is a wonderful and underutilized technique for discovering non-crashing design and implementation errors, too. Quite a few interesting bugs have been found by modifying the target programs to call abort() when say:

Two bignum libraries produce different outputs when given the same fuzzer-generated input.
An image library produces different outputs when asked to decode the same input image several times in a row.
A serialization/deserialization library fails to produce stable outputs when iteratively serializing and deserializing fuzzer-supplied data.
A compression library produces an output inconsistent with the input file when asked to compress and then decompress a particular blob.

Implementing these or similar sanity checks usually takes very little time; if you are the maintainer of a particular package, you can make this code conditional with #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION (a flag also shared with libfuzzer and honggfuzz) or #ifdef __AFL_COMPILER (this one is just for AFL++).

k) Known limitations & areas for improvement

Here are some of the most important caveats for AFL++:

AFL++ detects faults by checking for the first spawned process dying due to a signal (SIGSEGV, SIGABRT, etc.). Programs that install custom handlers for these signals may need to have the relevant code commented out. In the same vein, faults in child processes spawned by the fuzzed target may evade detection unless you manually add some code to catch that.
As with any other brute-force tool, the fuzzer offers limited coverage if encryption, checksums, cryptographic signatures, or compression are used to wholly wrap the actual data format to be tested.

To work around this, you can comment out the relevant checks (see utils/libpng_no_checksum/ for inspiration); if this is not possible, you can also write a postprocessor, one of the hooks of custom mutators. See /docs/custom_mutators/ on how to use AFL_CUSTOM_MUTATOR_LIBRARY.
There are some unfortunate trade-offs with ASAN and 64-bit binaries. This isn’t due to any specific fault of afl-fuzz.
There is no direct support for fuzzing network services, background daemons, or interactive apps that require UI interaction to work. You may need to make simple code changes to make them behave in a more traditional way. Preeny or libdesock may offer a relatively simple option, too - see: https://github.com/zardus/preeny or https://github.com/fkie-cad/libdesock

Some useful tips for modifying network-based services can be also found at: https://www.fastly.com/blog/how-to-fuzz-server-american-fuzzy-lop
Occasionally, sentient machines rise against their creators. If this happens to you, please consult https://lcamtuf.coredump.cx/prep/.

Beyond this, see /docs/install/ for platform-specific tips.

4. Triaging crashes

The coverage-based grouping of crashes usually produces a small data set that can be quickly triaged manually or with a very simple GDB or Valgrind script. Every crash is also traceable to its parent non-crashing test case in the queue, making it easier to diagnose faults.

Having said that, it’s important to acknowledge that some fuzzing crashes can be difficult to quickly evaluate for exploitability without a lot of debugging and code analysis work. To assist with this task, afl-fuzz supports a very unique “crash exploration” mode enabled with the -C flag.

In this mode, the fuzzer takes one or more crashing test cases as the input and uses its feedback-driven fuzzing strategies to very quickly enumerate all code paths that can be reached in the program while keeping it in the crashing state.

Mutations that do not result in a crash are rejected; so are any changes that do not affect the execution path.

The output is a small corpus of files that can be very rapidly examined to see what degree of control the attacker has over the faulting address, or whether it is possible to get past an initial out-of-bounds read - and see what lies beneath.

Oh, one more thing: for test case minimization, give afl-tmin a try. The tool can be operated in a very simple way:

./afl-tmin -i test_case -o minimized_result -- /path/to/program [...]

The tool works with crashing and non-crashing test cases alike. In the crash mode, it will happily accept instrumented and non-instrumented binaries. In the non-crashing mode, the minimizer relies on standard AFL++ instrumentation to make the file simpler without altering the execution path.

The minimizer accepts the -m, -t, -f, and @@ syntax in a manner compatible with afl-fuzz.

Another tool in AFL++ is the afl-analyze tool. It takes an input file, attempts to sequentially flip bytes and observes the behavior of the tested program. It then color-codes the input based on which sections appear to be critical and which are not; while not bulletproof, it can often offer quick insights into complex file formats.

casr-afl from CASR tools provides comfortable triaging for crashes found by AFL++. Reports are clustered and contain severity and other information.

casr-afl -i /path/to/afl/out/dir -o /path/to/casr/out/dir

5. CI fuzzing

Some notes on continuous integration (CI) fuzzing - this fuzzing is different to normal fuzzing campaigns as these are much shorter runnings.

If the queue in the CI is huge and/or the execution time is slow then you can also add AFL_NO_STARTUP_CALIBRATION=1 to skip the initial queue calibration phase and start fuzzing at once. But only do that if the calibration time is too long for your overall available fuzz run time.

Always:
- LTO has a much longer compile time which is diametrical to short fuzzing - hence use afl-clang-fast instead.
- If you compile with CMPLOG, then you can save compilation time and reuse that compiled target with the -c option and as the main fuzz target. This will impact the speed by ~15% though.
- AFL_FAST_CAL - enables fast calibration, this halves the time the saturated corpus needs to be loaded.
- AFL_CMPLOG_ONLY_NEW - only perform cmplog on new finds, not the initial corpus as this very likely has been done for them already.
- Keep the generated corpus, use afl-cmin and reuse it every time!
Additionally randomize the AFL++ compilation options, e.g.:
- 30% for AFL_LLVM_CMPLOG
- 5% for AFL_LLVM_LAF_ALL
Also randomize the afl-fuzz runtime options, e.g.:
- 65% for AFL_DISABLE_TRIM
- 50% for AFL_KEEP_TIMEOUTS
- 50% use a dictionary generated by AFL_LLVM_DICT2FILE + AFL_LLVM_DICT2FILE_NO_MAIN=1
- 10% use MOpt (-L 0)
- 40% for AFL_EXPAND_HAVOC_NOW
- 20% for old queue processing (-Z)
- for CMPLOG targets, 70% for -l 2, 10% for -l 3, 20% for -l 2AT
Do not run any -M modes, just running -S modes is better for CI fuzzing. -M enables old queue handling etc. which is good for a fuzzing campaign but not good for short CI runs.

How this can look like can, e.g., be seen at AFL++’s setup in Google’s oss-fuzz and clusterfuzz.

The End

Check out the FAQ. Maybe it answers your question (that you might not even have known you had ;-) ).

This is basically all you need to know to professionally run fuzzing campaigns. If you want to know more, the tons of texts in docs/ will have you covered.

Note that there are also a lot of tools out there that help fuzzing with AFL++ (some might be deprecated or unsupported), see /docs/third_party_tools/.