Abstract
This class is designed to introduce students to the best tools and technology available for automating vulnerability discovery and crash triage with a focus on delivering a practical approach to applying this technology in real deployments at any scale.
Through an applied understanding of introductory program analysis and binary translation, techniques for finding various bug classes and methods for improved crash debugging will be discussed. We will take a deep dive into fuzzing, covering all aspects of this practical approach to finding bugs. As the most approachable and versatile of the available tools, the student will apply various fuzzing techniques to several real-world pieces of software. Students will learn strategies for analyzing attack surface, writing grammars, and generating effective corpus. We will explore in detail the latest innovations such as harnessing code coverage for guided evolutionary fuzzing and symbolic reasoning for concolic fuzzing.
We approach crash analysis through the lens of scriptable debuggers and program analysis. We will apply tools like reverse debugging and memory debuggers to assist in interactively diagnosing root cause of crashes. Then we will leverage the power of dynamic taint tracking and graph slicing to help isolate the path of user controlled input in the program and identify the exact input bytes influencing a crash. Lastly, we will look at possible ways to aid in determining severity of a vulnerability.
This class will focus on x86/x64 architecture and target file parsers, network parsers, browsers, and kernel interfaces on both Windows and Linux environments.
Key Learning Objectives
- Learn an effective strategy for using the latest tools & technology to discover vulnerabilities
- Master the latest fuzzing techniques for file, network, browser, and kernel fuzzing
- Learn grammar fuzzing, evolutionary fuzzing, in-memory fuzzing, and symbolic fuzzing
- Best practices for corpus generation, fuzzer deployment, and targeting
- Leverage dynamic binary translation for efficient tracing and deep program inspection
- Learn how to leverage time travel debugging for crash triage on Linux and Windows
- Introduction to intermediate languages for program analysis
- Apply powerful techniques like taint analysis and graph slicing towards crash analysis
Who Should Attend
This class is meant for professional developers or security researchers looking to add an automation component to their software security analysis. Students wanting to learn a programmatic and tool driven approach to analyzing software vulnerabilities and crash triage will benefit from this course.
Pre-requisites
Students should be prepared to tackle challenging and diverse subject matter and be comfortable writing functions in in C/C++ and python to complete exercises involving completing plugins for the discussed platforms. Attendees should have basic experience with debugging native x86/x64 memory corruption vulnerabilities on Linux or Windows.
Hardware and Software Requirements
Students should have the latest VMware Player, Workstation, or Fusion working on their machine.
Agenda
Day 1:
- Analysis of generational and mutational fuzzing
- Attack surface analysis
- Effective mutation engines
- Effective corpus generation
- Protocol and file format grammars
- Crash detection
- Fuzzing file and network parsers with coverage guided fuzzing
- Fuzz any Ubuntu/Debian package with AFL
- Modifying targets and writing harnesses with LibFuzzer
- Fuzzing closed source parsers with QEMU and Dyninst
- Best practices for high performance fuzzing
- System configuration
- Corpus generation techniques
- Cross-fuzzing difficult parsers
Day 2:
- Dynamic Binary Translation for Fuzzing and Triage
- Effectively instrument Linux and Windows with binary translation
- Introduction to Valgrind, Dr. Memory, and Address Sanitizer
- Introduction to PIN, DynamoRIO, and Dyninst internals
- Identifying hook locations with Debuggers and DBI
- Fuzzing parsers with WinAFL
- Optimizing harnesses for exported APIs
- Hooking closed source command line applications
- Deep hooks into private library functions with global state
- Fuzzing internal data streams in complex OLE objects
Day 3:
- Time Travel Debugging
- Introduction to time travel debugging
- Crash analysis with reverse debugging on Linux
- Crash analysis with reverse debugging on Windows
- Taint assisted root cause analysis
- Introduction to dynamic taint analysis
- Taint slicing for for root cause analysis
- Fuzzing browsers with evolutionary grammar fuzzing
- Understanding grammars and object models
- Fuzzing object models with dynamic grammar fuzzing
- Improving grammar fuzzers with feedback metrics
Day 4:
- Fuzzing Kernels
- Fuzzing system calls with type aware API fuzzing
- Evolutionary system call fuzzing
- Corpus collection via API hooking
- Fuzzing IOCTLs with coverage guided kernel fuzzing
- Symbolic and Concolic Execution
- Introduction to constraint solving
- Reversing driver interfaces with symbolic execution
- Forward symbolic execution for exploitability analysis
- Concolic execution for test case generation