FLARE Script Series: Reverse Engineering WebAssembly Modules Using the idawasm IDA Pro Plugin

Introduction

This post continues the FireEye Labs Advanced Reverse Engineering
(FLARE) script series. Here, we introduce idawasm, an IDA Pro plugin
that provides a loader and processor modules for WebAssembly modules.
idawasm works on all operating systems supported by IDA Pro, and can
be obtained from the idawasm GitHub project.

Motivation

You may have competed in this year’s Flare-On
challenge
and reached the fifth level only to discover a new file
format: a WebAssembly (“wasm”) module. In order to proceed, you had to
reverse engineer the key check logic contained in this binary file for
the WebAssembly stack-based virtual machine. But first, you probably
learned a bit about wasm:

"Wasm is designed as a portable
target for compilation of high-level languages like C/C++/Rust,
enabling deployment on the web for client and server applications.
The Wasm stack machine is
designed to be encoded in a size- and load-time-efficient binary
format
. WebAssembly aims to execute at native speed by taking
advantage of common
hardware capabilities
available on a wide range of platforms."

via: WebAssembly

While WebAssembly is a fairly new standard, some tools for analysis
already exist:

  • The WebAssembly Binary
    Toolkit
     provides the command line tool wasm2wat that translates
    a .wasm file into the human-readable .wat format, as well as a basic
    decompiler called wasm2c.
  • The web-based WebAssembly Studio
    IDE
     exposes features for extracting .wasm files into interesting
    formats, including translations to x86-64 produced by the Firefox
    SpiderMonkey JIT engine.
  • Radare2 can
    disassemble instructions, though it doesn’t reconstruct control flow
    graphs just yet.

Unfortunately, few of these tools are interactive or familiar to us.
If we could analyze .wasm files in IDA Pro, like we do with .exe and
.so files, then we might be better prepared for malware distributed as WebAssembly.

The idawasm IDA
Pro plugin is a loader and processor module that enables analysts to
review WebAssembly modules using a familiar interface. Let’s take a
peek at some of its major features.

Whirlwind Tour of idawasm

Once you’ve installed idawasm, you can load .wasm files into IDA
Pro. The "Load a new file" dialog will indicate when the
loader module recognizes a WebAssembly module. Currently, the plugin
supports the MVP (version 1) release of WebAssembly, as shown in
Figure 1.



Figure 1: IDA Pro loading a WebAssembly module

The processor module then gets to work and reconstructs the control
flow, which enables IDA’s graph mode. This makes it easy to identify
high-level control constructs, such as if statements and while loops.
On the FLARE team, we’ve found that when a C program is compiled for
WebAssembly, the resulting control flow graphs mirror those of an x86
compilation target. Although the specific instructions differ, our
techniques and intuitions we’ve learned for x86 apply to wasm too.
Note how easy it is to identify the pair of if statements in Figure 2.



Figure 2: Control flow graph of
WebAssembly instructions

In addition to control flow, the processor module parses and renders
function names and types using metadata embedded with .wasm files. It
also extracts cross-references to global variables, enabling an
analyst to “List cross-references” and find code that manipulates
interesting values. Of course, all of the elements that idawasm parses
can be interactively renamed and commented. Therefore, you can record
knowledge about WebAssembly malware samples in .idb files and share
them with other analysts.

For example, Figure 3 shows how an analyst has renamed local
variables and added comments that explain how the function frame is
manipulated during a function prologue.



Figure 3: Annotated WebAssembly in IDA Pro

When the idawasm processor detects that a WebAssembly module was
compiled by LLVM, it enables enhanced analysis. LLVM uses primitives
provided by the WebAssembly specification to implement a function
frame stack in global memory used for mutable function-local
variables. In the function prologue, the first few instructions
allocate the frame on this global stack, while the epilogue cleans it
up prior to return. idawasm automatically finds references to the
global frame stack pointer and reconstructs the frame layout of each
function. Using this information, the processor updates IDA’s stack
frame structures and marks immediate constants as offsets into these
structures. This means that many more instruction operands are
actually symbols that can be commented and renamed, as shown in Figure 4.



Figure 4: Automated reconstruction of
function frame

By convention, WebAssembly compilers emit instructions that
reference variables in Single Static Assignment (SSA) form. This is a
benefit to browser engines, because code that is already in SSA form
is easier to import into analysis systems such as optimizers and JIT
engines. Using SSA form is why even the simplest functions have dozens
or hundreds of local variables.

But as a human, SSA form is tedious to analyze since we must trace
operations across many separate local variables. To help us out,
idawasm includes a WebAssembly emulator that can track instructions at
a symbolic level. This enables us to collapse a sequence of
simple-but-related instructions into a single complex expression.

To use it, we select a region of instructions within a single basic
block and run the wasm_emu.py script. The script emulates the
instructions, simplifies their effects, and renders the effect to
global variables, locals, memory, and the stack. Figure 5 shows how a
function’s prologue is simplified to a single global variable update:



Figure 5: wasm_emu.py show effects of
function frame allocation

When there are many arithmetic operations, wasm_emu.py often
simplifies into a single meaningful expression. For example, Figure 6
shows how 32 instructions boil down into a single XOR across two input bytes:



Figure 6: wasm_emu.py simplifying many
instructions to a complex instruction

Conclusion

idawasm is an IDA Pro plugin that enables support for WebAssembly
modules via a loader and processor. This lets analysts use a familiar
interface to reverse engineer .wasm files and share their work with
others. The wasm_emu.py script can also be helpful for understanding
the effects of long streams of WebAssembly instructions. Now dealing
with this new file format and architecture won’t be as scary, so head
on over to the idawasm GitHub
project
to start using it!



*** This is a Security Bloggers Network syndicated blog from Threat Research authored by Threat Research Blog. Read the original post at: http://www.fireeye.com/blog/threat-research/2018/10/reverse-engineering-webassembly-modules-using-the-idawasm-ida-pro-plugin.html