A symbolic-model-guided fuzzer for TLS

Overview

tlspuffin

Logo with Penguin

TLS Protocol Under FuzzINg
A symbolic-model-guided fuzzer for TLS

Disclaimer: The term "symbolic-model-guided" should not be confused with symbolic execution or concolic fuzzing.

Description

Fuzzing implementations of cryptographic protocols is challenging. In contrast to traditional fuzzing of file formats, cryptographic protocols require a specific flow of cryptographic and mutually dependent messages to reach deep protocol states. The specification of the TLS protocol describes sound flows of messages and cryptographic operations.

Although the specification has been formally verified multiple times with significant results, a gap has emerged from the fact that implementations of the same protocol have not undergone the same logical analysis. Because the development of cryptographic protocols is error-prone, multiple security vulnerabilities have already been discovered in implementations in TLS which are not present in its specification.

Inspired by symbolic protocol verification, we present a reference implementation of a fuzzer named tlspuffin which employs a concrete semantic to execute TLS 1.2 and 1.3 symbolic traces. In fact attacks which mix \TLS versions are in scope of this implementation. This method allows us to utilize a genetic fuzzing algorithm to fuzz protocol flows, which is described by the following three stages.

  • By mutating traces we can deviate from the specification to test logical flaws.
  • Selection of interesting protocol flows advance the fuzzing procedure.
  • A security violation oracle supervises executions for the absence of vulnerabilities.

The novel approach allows rediscovering known vulnerabilities, which are out-of-scope for classical bit-level fuzzers. This proves that it is capable of reaching critical protocol states. In contrast to the promising methodology no new vulnerabilities were found by tlspuffin. This can can be explained by the fact that the implementation effort of TLS protocol primitives and extensions is high and not all features of the specification have been implemented. Nonetheless, the innovating approach is promising in terms of quickly reaching high edge coverage, expressiveness of executable protocol traces and stable and extensible implementation.

Features

  • Uses the LibAFL fuzzing framework
  • Fuzzer which is inspired by the Dolev-Yao symbolic model used in protocol verification
  • Domain specific mutators for Protocol Fuzzing!
  • Supported Libraries Under Test: OpenSSL 1.0.1f, 1.0.2u, 1.1.1k and LibreSSL 3.3.3
  • Reproducible for each LUT. We use Git submodules to link to forks this are in the tlspuffin organisation
  • 70% Test Coverage
  • Writtin in Rust!

Building

Now, to build the project:

git clone [email protected]/tlspuffin/tlspuffin
git submodule update --init --recursive
cargo build

Running

Fuzz using three clients:

cargo run --bin tlspuffin -- --cores 0-3

Note: After switching the Library Under Test or its version do a clean rebuild (cargo clean). For example when switching from OpenSSL 1.0.1 to 1.1.1.

Testing

cargo test

Command-line Interface

The syntax for the command-line of is:

      tlspuffin [⟨options] [⟨sub-commands⟩]

Global Options

Before we explain each sub-command, we first go over the options in the following.

  • -c, --cores ⟨spec⟩

    This option specifies on which cores the fuzzer should assign its worker processes. It can either be specified as a list by using commas "0,1,2,7" or as a range "0-7". By default, it runs just on core 0.

  • -i, --max-iters ⟨i⟩

    This option allows to bound the amount of iterations the fuzzer does. If omitted, then infinite iterations are done.

  • -p, --port ⟨n⟩

    As specified in [sec:design-multiprocessing] the initial communication between the fuzzer broker and workers happens over TCP/IP. Therefore, the broker requires a port allocation. The default port is 1337.

  • -s, --seed ⟨n⟩

    Defines an initial seed for the prng used for mutations. Note that this does not make the fuzzing deterministic, because of randomness introduced by the multiprocessing (see [sec:design-multiprocessing]).

Sub-commands

Now we will go over the sub-commands execute, plot, experiment, and seed.

  • execute ⟨input⟩

    This sub-command executes a single trace persisted in a file. The path to the file is provided by the ⟨input⟩ argument.

  • plot ⟨input⟩ ⟨format⟩ ⟨output_prefix⟩

    This sub-command plots the trace stored at ⟨input⟩ in the format specified by ⟨format⟩. The created graphics are stored at a path provided by ⟨output_prefix⟩. The option --multiple can be provided to create for each step in the trace a separate file. If the option --tree is given, then only a single graphic which contains all steps is produced.

  • experiment

    This sub-command initiates an experiment. Experiments are stored in a directory named experiments/ in the current working directory. An experiment consists of a directory which contains . The title and description of the experiment can be specified with --title ⟨t⟩ and --description ⟨d⟩ respectively. Both strings are persisted in the metadata of the experiment, together with the current commit hash of , the version and the current date and time.

  • seed

    This sub-command serializes the default seed corpus in a directory named corpus/ in the current working directory. The default corpus is defined in the source code of using the trace dsl.

Rust Setup

Install rustup.

The toolchain will be automatically downloaded when building this project. See ./rust-toolchain.toml for more details about the toolchain.

Make sure that you have the clang compiler installed. Optionally, also install llvm to have additional tools like sancov available. Also make sure that you have the usual tools for building it like make, gcc etc. installed. They may be needed to build OpenSSL.

Advanced Features

Running with ASAN

ASAN_OPTIONS=abort_on_error=1 \
    cargo run --bin tlspuffin --features asan -- --cores 0-3

It is important to enable abort_on_error, else the fuzzer workers fail to restart on crashes.

Generate Corpus Seeds

cargo run --bin tlspuffin -- seed

Plot Symbolic Traces

To plot SVGs do the following:

cargo run --bin tlspuffin -- plot corpus/seed_client_attacker12.trace svg ./plots/seed_client_attacker12

Note: This requires that the dot binary is in on your path. Note: The utility tools/plot-corpus.sh plots a whole directory

Execute a Symbolic Trace (with ASAN)

To analyze crashes you can also execute a trace which crashes the testing harness using ASAN:

cargo run --bin tlspuffin -- execute test.trace

To do the same with ASAN enabled:

ASAN_OPTIONS=detect_leaks=0 \
      cargo run --bin tlspuffin --features asan -- execute test.trace

Crash Deduplication

Creates log files for each crash and parses ASAN crashes to group crashes together.

tools/analyze-crashes.sh

Benchmarking

There is a benchmark which compares the execution of the dynamic functions to directly executing them in benchmark.rs. You can run them using:

cargo bench
xdg-open target/criterion/report/index.html

Documentation

This generates the documentation for this crate and opens the browser. This also includes the documentation of every dependency like LibAFL or rustls.

cargo doc --open

You can also view the up-to-date documentation here.

You might also like...
Theano is a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. It can use GPUs and perform efficient symbolic differentiation.

============================================================================================================ `MILA will stop developing Theano https:

ATOMIC 2020: On Symbolic and Neural Commonsense Knowledge Graphs
ATOMIC 2020: On Symbolic and Neural Commonsense Knowledge Graphs

(Comet-) ATOMIC 2020: On Symbolic and Neural Commonsense Knowledge Graphs Paper Jena D. Hwang, Chandra Bhagavatula, Ronan Le Bras, Jeff Da, Keisuke Sa

PIGLeT: Language Grounding Through Neuro-Symbolic Interaction in a 3D World [ACL 2021]

piglet PIGLeT: Language Grounding Through Neuro-Symbolic Interaction in a 3D World [ACL 2021] This repo contains code and data for PIGLeT. If you like

Symbolic Parallel Adaptive Importance Sampling for Probabilistic Program Analysis in JAX

SYMPAIS: Symbolic Parallel Adaptive Importance Sampling for Probabilistic Program Analysis Overview | Installation | Documentation | Examples | Notebo

Official repository for the paper, MidiBERT-Piano: Large-scale Pre-training for Symbolic Music Understanding.
Official repository for the paper, MidiBERT-Piano: Large-scale Pre-training for Symbolic Music Understanding.

MidiBERT-Piano Authors: Yi-Hui (Sophia) Chou, I-Chun (Bronwin) Chen Introduction This is the official repository for the paper, MidiBERT-Piano: Large-

Source code and Dataset creation for the paper "Neural Symbolic Regression That Scales"

NeuralSymbolicRegressionThatScales Pytorch implementation and pretrained models for the paper "Neural Symbolic Regression That Scales", presented at I

Data and Code for ACL 2021 Paper
Data and Code for ACL 2021 Paper "Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoning"

Introduction Code and data for ACL 2021 Paper "Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoning". We cons

PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and reinforcement learning
PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and reinforcement learning

safe-control-gym Physics-based CartPole and Quadrotor Gym environments (using PyBullet) with symbolic a priori dynamics (using CasADi) for learning-ba

Driller: augmenting AFL with symbolic execution!

Driller Driller is an implementation of the driller paper. This implementation was built on top of AFL with angr being used as a symbolic tracer. Dril

Comments
  • Support for WolfSSL through a new rust-wolfssl crate

    Support for WolfSSL through a new rust-wolfssl crate

    I tried another approach as I struggled with #136 to debug some SEGFAULT.

    Here the approach is to clone rust-openssl, minimize the exposed interface while exposing what we need for wolfssl_binding.rs, then plug in wolfssl-sys instead of openssl-sys.

    opened by LCBH 3
  • Rediscover wolfSSL vulnerabilities

    Rediscover wolfSSL vulnerabilities

    • CVE-2020-12457 in <4.5.0 https://nvd.nist.gov/vuln/detail/CVE-2020-12457 (DDOS against server, needs change_cipher_spec (CCS) message mutations)
    • CVE 2020-24613 in <4.5.0 https://nvd.nist.gov/vuln/detail/CVE-2020-24613 in < 4.5.0, TLS 1.3 server auth. bypass
    • CVE-2021-3336 in < 4.7.0 https://nvd.nist.gov/vuln/detail/CVE-2021-3336, TLS 1.3 server auth. bypass
    • CVE 2022-25638 in < 5.2.0 https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-25638, TLS 1.3 server auth. bypass
    • CVE-2022-25640 in <5.2.0 https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-25640 (attack on client authentication, needs client authentication through cert.)
    opened by maxammann 1
Releases(evaluation)
Training vision models with full-batch gradient descent and regularization

Stochastic Training is Not Necessary for Generalization -- Training competitive vision models without stochasticity This repository implements trainin

Jonas Geiping 32 Jan 06, 2023
[ICML 2021] DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning | 斗地主AI

[ICML 2021] DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning DouZero is a reinforcement learning framework for DouDizhu (斗地主), t

Kwai Inc. 3.1k Jan 04, 2023
Orbivator AI - To Determine which features of data (measurements) are most important for diagnosing breast cancer and find out if breast cancer occurs or not.

Orbivator_AI Breast Cancer Wisconsin (Diagnostic) GOAL To Determine which features of data (measurements) are most important for diagnosing breast can

anurag kumar singh 1 Jan 02, 2022
Many Class Activation Map methods implemented in Pytorch for CNNs and Vision Transformers. Including Grad-CAM, Grad-CAM++, Score-CAM, Ablation-CAM and XGrad-CAM

Class Activation Map methods implemented in Pytorch pip install grad-cam ⭐ Tested on many Common CNN Networks and Vision Transformers. ⭐ Includes smoo

Jacob Gildenblat 6.6k Jan 06, 2023
Get a Grip! - A robotic system for remote clinical environments.

Get a Grip! Within clinical environments, sterilization is an essential procedure for disinfecting surgical and medical instruments. For our engineeri

Jay Sharma 1 Jan 05, 2022
Interacting Two-Hand 3D Pose and Shape Reconstruction from Single Color Image (ICCV 2021)

Interacting Two-Hand 3D Pose and Shape Reconstruction from Single Color Image Interacting Two-Hand 3D Pose and Shape Reconstruction from Single Color

75 Dec 02, 2022
NCVX (NonConVeX): A User-Friendly and Scalable Package for Nonconvex Optimization in Machine Learning.

The source code is temporariy removed, as we are solving potential copyright and license issues with GRANSO (http://www.timmitchell.com/software/GRANS

SUN Group @ UMN 28 Aug 03, 2022
BanditPAM: Almost Linear-Time k-Medoids Clustering

BanditPAM: Almost Linear-Time k-Medoids Clustering This repo contains a high-performance implementation of BanditPAM from BanditPAM: Almost Linear-Tim

254 Dec 12, 2022
DeepConsensus uses gap-aware sequence transformers to correct errors in Pacific Biosciences (PacBio) Circular Consensus Sequencing (CCS) data.

DeepConsensus DeepConsensus uses gap-aware sequence transformers to correct errors in Pacific Biosciences (PacBio) Circular Consensus Sequencing (CCS)

Google 149 Dec 19, 2022
TensorFlow implementation of the paper "Hierarchical Attention Networks for Document Classification"

Hierarchical Attention Networks for Document Classification This is an implementation of the paper Hierarchical Attention Networks for Document Classi

Quoc-Tuan Truong 83 Dec 05, 2022
Mosaic of Object-centric Images as Scene-centric Images (MosaicOS) for long-tailed object detection and instance segmentation.

MosaicOS Mosaic of Object-centric Images as Scene-centric Images (MosaicOS) for long-tailed object detection and instance segmentation. Introduction M

Cheng Zhang 27 Oct 12, 2022
HyperCube: Implicit Field Representations of Voxelized 3D Models

HyperCube: Implicit Field Representations of Voxelized 3D Models Authors: Magdalena Proszewska, Marcin Mazur, Tomasz Trzcinski, Przemysław Spurek [Pap

Magdalena Proszewska 3 Mar 09, 2022
Real life contra a deep learning project built using mediapipe and openc

real-life-contra Description A python script that translates the body movement into in game control. Welcome to all new real life contra a deep learni

Programminghut 7 Jan 26, 2022
Galileo library for large scale graph training by JD

近年来,图计算在搜索、推荐和风控等场景中获得显著的效果,但也面临超大规模异构图训练,与现有的深度学习框架Tensorflow和PyTorch结合等难题。 Galileo(伽利略)是一个图深度学习框架,具备超大规模、易使用、易扩展、高性能、双后端等优点,旨在解决超大规模图算法在工业级场景的落地难题,提

JD Galileo Team 128 Nov 29, 2022
Language Used: Python . Made in Jupyter(Anaconda) notebook.

FACE-DETECTION-ATTENDENCE-SYSTEM Made in Jupyter(Anaconda) notebook. Language Used: Python Steps to perform before running the program : Install Anaco

1 Jan 12, 2022
Securetar - A streaming wrapper around python tarfile and allow secure handling files and support encryption

Secure Tar Secure Tarfile library It's a streaming wrapper around python tarfile

Pascal Vizeli 2 Dec 09, 2022
Official Code Release for Container : Context Aggregation Network

Container: Context Aggregation Network Official Code Release for Container : Context Aggregation Network Comparion between CNN, MLP-Mixer and Transfor

peng gao 42 Nov 17, 2021
Bayesian optimisation library developped by Huawei Noah's Ark Library

Bayesian Optimisation Research This directory contains official implementations for Bayesian optimisation works developped by Huawei R&D, Noah's Ark L

HUAWEI Noah's Ark Lab 395 Dec 30, 2022
PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).

PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR)

Ilya Kostrikov 3k Dec 31, 2022
Using Convolutional Neural Networks (CNN) for Semantic Segmentation of Breast Cancer Lesions (BRCA)

Using Convolutional Neural Networks (CNN) for Semantic Segmentation of Breast Cancer Lesions (BRCA). Master's thesis documents. Bibliography, experiments and reports.

Erick Cobos 73 Dec 04, 2022