Awesome
libprotobuf-mutator
Overview
libprotobuf-mutator is a library to randomly mutate protobuffers. <BR> It could be used together with guided fuzzing engines, such as libFuzzer.
Quick start on Debian/Ubuntu
Install prerequisites:
sudo apt-get update
sudo apt-get install protobuf-compiler libprotobuf-dev binutils cmake \
ninja-build liblzma-dev libz-dev pkg-config autoconf libtool
Compile and test everything:
mkdir build
cd build
cmake .. -GNinja -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Debug
ninja check
Clang is only needed for libFuzzer integration. <BR>
By default, the system-installed version of
protobuf is used. However, on some
systems, the system version is too old. You can pass
LIB_PROTO_MUTATOR_DOWNLOAD_PROTOBUF=ON
to cmake to automatically download and
build a working version of protobuf.
Installation:
ninja
sudo ninja install
This installs the headers, pkg-config, and static library.
By default the headers are put in /usr/local/include/libprotobuf-mutator
.
Usage
To use libprotobuf-mutator simply include mutator.h and mutator.cc into your build files.
The ProtobufMutator
class implements mutations of the protobuf
tree structure and mutations of individual fields.
The field mutation logic is very basic --
for better results you should override the ProtobufMutator::Mutate*
methods with more sophisticated logic, e.g.
using libFuzzer's mutators.
To apply one mutation to a protobuf object do the following:
class MyProtobufMutator : public protobuf_mutator::Mutator {
public:
// Optionally redefine the Mutate* methods to perform more sophisticated mutations.
}
void Mutate(MyMessage* message) {
MyProtobufMutator mutator;
mutator.Seed(my_random_seed);
mutator.Mutate(message, 200);
}
See also the ProtobufMutatorMessagesTest.UsageExample
test from
mutator_test.cc.
Integrating with libFuzzer
LibFuzzerProtobufMutator can help to integrate with libFuzzer. For example
#include "src/libfuzzer/libfuzzer_macro.h"
DEFINE_PROTO_FUZZER(const MyMessageType& input) {
// Code which needs to be fuzzed.
ConsumeMyMessageType(input);
}
Please see libfuzzer_example.cc as an example.
Mutation post-processing (experimental)
Sometimes it's necessary to keep particular values in some fields without which the proto is going to be rejected by fuzzed code. E.g. code may expect consistency between some fields or it may use some fields as checksums. Such constraints are going to be significant bottleneck for fuzzer even if it's capable of inserting acceptable values with time.
PostProcessorRegistration can be used to avoid such issue and guide your fuzzer towards interesting code. It registers callback which will be called for each message of particular type after each mutation.
static protobuf_mutator::libfuzzer::PostProcessorRegistration<MyMessageType> reg = {
[](MyMessageType* message, unsigned int seed) {
TweakMyMessage(message, seed);
}};
DEFINE_PROTO_FUZZER(const MyMessageType& input) {
// Code which needs to be fuzzed.
ConsumeMyMessageType(input);
}
Optional: Use seed if callback uses random numbers. It may help later with debugging.
Important: Callbacks should be deterministic and avoid modifying good messages. Callbacks are called for both: mutator generated and user provided inputs, like corpus or bug reproducer. So if callback performs unnecessary transformation it may corrupt the reproducer so it stops triggering the bug.
Note: You can add callback for any nested message and you can add multiple callbacks for the same message type.
static PostProcessorRegistration<MyMessageType> reg1 = {
[](MyMessageType* message, unsigned int seed) {
TweakMyMessage(message, seed);
}};
static PostProcessorRegistration<MyMessageType> reg2 = {
[](MyMessageType* message, unsigned int seed) {
DifferentTweakMyMessage(message, seed);
}};
static PostProcessorRegistration<MyMessageType::Nested> reg_nested = {
[](MyMessageType::Nested* message, unsigned int seed) {
TweakMyNestedMessage(message, seed);
}};
DEFINE_PROTO_FUZZER(const MyMessageType& input) {
// Code which needs to be fuzzed.
ConsumeMyMessageType(input);
}
UTF-8 strings
"proto2" and "proto3" handle invalid UTF-8 strings differently. In both cases string should be UTF-8, however only "proto3" enforces that. So if fuzzer is applied to "proto2" type libprotobuf-mutator will generate any strings including invalid UTF-8. If it's a "proto3" message type, only valid UTF-8 will be used.
Extensions
Currently the library does not mutate extensions. This can be a problem if extension contains required fields so the library will not be able to change the message into valid initialized state. You can use post processing hooks to cleanup/initialize the message as workaround.
Users of the library
Grammars
- GIF, https://github.com/google/oss-fuzz/tree/master/projects/giflib
- JSON
- Lua 5.1 Language,
- PNG, https://github.com/google/oss-fuzz/tree/master/projects/libpng-proto
- SQL
- Solidity Language, https://github.com/ethereum/solidity/tree/develop/test/tools/ossfuzz
- XML
- JPEG, https://source.chromium.org/chromium/chromium/src/+/main:media/gpu/vaapi/fuzzers/jpeg_decoder/
Bugs found with help of the library
Chromium
- AppCache exploit (Actual still restricted bug)
- Stack Buffer Overflow in QuicClientPromisedInfo
- null dereference in sqlite3ExprCompare
Envoy
- strftime overflow
- Heap-use-after-free in Envoy::Upstream::SubsetLoadBalancer::updateFallbackSubset
- Heap-use-after-free in Envoy::Secret::SecretManagerImpl
- Heap-buffer-overflow in Envoy::Http::HeaderString