barf-project

BARF : Binary Analysis and Reverse engineering Framework

The analysis of binary code is a crucial activity in many areas of the computer sciences and software engineering disciplines ranging from software security and program analysis to reverse engineering. Manual binary analysis is a difficult and time-consuming task and there are software tools that seek to automate or assist human analysts. However, most of these tools have several technical and commercial restrictions that limit access and use by a large portion of the academic and practitioner communities. BARF is an open source binary analysis framework that aims to support a wide range of binary code analysis tasks that are common in the information security discipline. It is a scriptable platform that supports instruction lifting from multiple architectures, binary translation to an intermediate representation, an extensible framework for code analysis plugins and interoperation with external tools such as debuggers, SMT solvers and instrumentation tools. The framework is designed primarily for human-assisted analysis but it can be fully automated.

The BARF project includes BARF and related tools and packages. So far the project is composed of the following items:

• BARF : A multiplatform open source Binary Analysis and Reverse engineering Framework
• PyAsmJIT : A JIT for the Intel x86_64 and ARM architecture.
• Tools built upon BARF:
  • BARFgadgets : Lets you search, classifiy and verify ROP gadgets inside a binary program.
  • BARFcfg : Lets you recover the control-flow graph of the functions of a binary program.
  • BARFcg : Lets you recover the call graph of the functions of a binary program.

For more information, see:

• BARF: A multiplatform open source Binary Analysis and Reverse engineering Framework (Whitepaper) [en]
• BARFing Gadgets (ekoparty2014 presentation) [es]

Current status:

Latest Release v0.6.0 URL https://github.com/programa-stic/barf-project/releases/tag/v0.6.0 Change Log https://github.com/programa-stic/barf-project/blob/v0.6.0/CHANGELOG.md

All packages were tested on Ubuntu 16.04 (x86_64).

BARF

BARF is a Python package for binary analysis and reverse engineering. It can:

• Load binary programs in different formats (ELF, PE, etc),
• It supports the Intel x86 architecture for 32 and 64 bits,
• It supports the ARM architecture for 32 bits,
• It operates on an intermediate language (REIL) thus all analysis algorithm are architecture-agnostic,
• It has integration with Z3 and CVC4 SMT solvers which means that you can express fragments of code as formulae and check restrictions on them.

It is currently under development.

Installation

BARF depends on the following SMT solvers:

• Z3 : A high-performance theorem prover being developed at Microsoft Research.
• CVC4 : An efficient open-source automatic theorem prover for satisfiability modulo theories (SMT) problems.

The following command installs BARF on your system:

$ sudo python setup.py install

You can also install it locally:

$ sudo python setup.py install --user

Notes

• Only one SMT solver is needed in order to work. You may choose between Z3 and CVC4 or install both.
• To run some tests you need to install PyAsmJIT first: sudo pip install pyasmjit
• You may need to install Graphviz: sudo apt-get install graphviz

Quickstart

This is a very simple example which shows how to open a binary file and print each instruction with its translation to the intermediate language (REIL).

from barf import BARF

# Open binary file.
barf = BARF("examples/misc/samples/bin/branch4.x86")

# Print assembly instruction.
for addr, asm_instr, reil_instrs in barf.translate():
    print("{:#x} {}".format(addr, asm_instr))

    # Print REIL translation.
    for reil_instr in reil_instrs:
        print("\t{}".format(reil_instr))

We can also recover the CFG and save it to a .dot file.

# Recover CFG.
cfg = barf.recover_cfg()

# Save CFG to a .dot file.
cfg.save("branch4.x86_cfg")

We can check restrictions on code using a SMT solver. For instance, suppose you have the following code:

…

And you want to know what values you have to assign to memory locations ebp-0x4, ebp-0x8 and ebp-0xc in order to obtain a specific value in eax register after executing the code.

First, we add the instructions to the analyzer component.

from barf import BARF

# Open ELF file
barf = BARF("examples/misc/samples/bin/constraint1.x86")

# Add instructions to analyze.
for addr, asm_instr, reil_instrs in barf.translate(0x80483ed, 0x8048401):
    for reil_instr in reil_instrs:
        barf.code_analyzer.add_instruction(reil_instr)

Then, we generate expressions for each variable of interest and add the desired restrictions on them.

ebp = barf.code_analyzer.get_register_expr("ebp", mode="post")

# Preconditions: set range for variable a and b
a = barf.code_analyzer.get_memory_expr(ebp-0x8, 4, mode="pre")
b = barf.code_analyzer.get_memory_expr(ebp-0xc, 4, mode="pre")

for constr in [a >= 2, a <= 100, b >= 2, b <= 100]:
    barf.code_analyzer.add_constraint(constr)

# Postconditions: set desired value for the result
c = barf.code_analyzer.get_memory_expr(ebp-0x4, 4, mode="post")

for constr in [c >= 26, c <= 28]:
    barf.code_analyzer.add_constraint(constr)

Finally, we check is the restrictions we establish can be resolved.

if barf.code_analyzer.check() == 'sat':
    print("[+] Satisfiable! Possible assignments:")

    # Get concrete value for expressions
    a_val = barf.code_analyzer.get_expr_value(a)
    b_val = barf.code_analyzer.get_expr_value(b)
    c_val = barf.code_analyzer.get_expr_value(c)

    # Print values
    print("- a: {0:#010x} ({0})".format(a_val))
    print("- b: {0:#010x} ({0})".format(b_val))
    print("- c: {0:#010x} ({0})".format(c_val))

    assert a_val + b_val + 5 == c_val
else:
    print("[-] Unsatisfiable!")

You can see these and more examples in the examples directory.

Overview

The framework is divided in three main components: core, arch and analysis.

Core

This component contains essential modules:

• REIL: Provides definitions for the REIL language. It, also, implements an emulator and a parser.
• SMT: Provides means to interface with Z3 and CVC4 SMT solver. Also, it provides functionality to translate REIL instructions to SMT expressions.
• BI: The Binary Interface module is responsible for loading binary files for processing (it uses PEFile and PyELFTools.)

Arch

Each supported architecture is provided as a subcomponent which contains the following modules.

• Architecture: Describes the architecture, i.e., registers, memory address size.
• Translator: Provides translators to REIL for each supported instruction.
• Disassembler: Provides disassembling functionalities (it uses Capstone.)
• Parser: Transforms instruction from string to object form.

Analysis

So far this component consists of modules: Control-Flow Graph, Call Graph and Code Analyzer. The first two, provides functionality for CFG and CG recovery, respectively. The latter, its a high-level interface to the SMT solver related functionality.

Tools

BARFgadgets

BARFgadgets is a Python script built upon BARF that lets you search, classifiy and verify ROP gadgets inside a binary program. The search stage finds all ret-, jmp- and call-ended gadgets inside the binary. The classification stage classifies previously found gadgets according to the following types:

• No-Operation,
• Move Register,
• Load Constant,
• Arithmetic/Logical Operation,
• Load Memory,
• Store Memory,
• Arithmetic/Logical Load,
• Arithmetic/Logical Store and
• Undefined.

This is done through instruction emulation. Finally, the verification stage consists of using a SMT solver to verify the semantic assigned to each gadget in the second stage.

…

For more information, see README.

BARFcfg

BARFcfg is a Python script built upon BARF that lets you recover the control-flow graph of a binary program.

…

BARFcg

BARFcg is a Python script built upon BARF that lets you recover the call graph of a binary program.

…

PyAsmJIT

PyAsmJIT is a Python package for x86_64/ARM assembly code generation and execution.

This package was developed in order to test BARF instruction translation from x86_64/ARM to REIL. The main idea is to be able to run fragments of code natively. Then, the same fragment is translated to REIL and executed in a REIL VM. Finally, both final contexts (the one obtained through native execution and the one from emulation) are compare for differences.

For more information, see PyAsmJIT.

License

The BSD 2-Clause License. For more information, see LICENSE.