Provides support for using zelos (a python-based binary instrumentation and emulation platform) as a concrete execution engine in angr via the symbion interface. Symbion provides an interface that enables angr to get and set program state from an external execution engine. Conversely, this enables zelos to take advantage of the symbolic execution capabilities of angr.

You first need angr with symbion. Their documentation recommends installation in a separate virtual environment, as several python packages are customized, including unicorn. This is the boilerplate to setup angr in a new virtual environment. Refer to their documentation for more comprehensive instructions.

$ sudo apt-get install python3-dev libffi-dev build-essential cmake gdb-multiarch python3-venv
$ git clone https://github.com/angr/angr-targets.git

$ python3 -m venv ~/.venv/angr && source ~/.venv/angr/bin/activate
(angr) $ pip install wheel && pip install angr && pip install -e angr-targets

Once you have the angr environment setup, install the angr-zelos-target package within the environment to enable zelos as a concrete target:

(angr) $ pip install angr-zelos-target

Install the zelos emulator in a separate virtual environment, e.g.:

(zelos) $ pip install zelos

Write your angr scripts as you usually would, but specify the zelos concrete target when creating the project:

from angr_zelos_target import ZelosConcreteTarget

...

zelos_target = ZelosConcreteTarget()
project = angr.Project(
    filepath,
    concrete_target=zelos_target,
    use_sim_procedures=True,
)

Use the angr.exploration_techniques.Symbion exploration technique when you want to concretely execution in zelos.

Before running your angr script, start the zelos zdbserver on the target binary:

(zelos) $ python -m zelos.tools.zdbserver FILENAME

The zdbserver and angr scripts can run on the same system, but be sure to run them in separate python environments, as both packages use different versions of the unicorn CPU emulator.

As an example, we have reimplemented the symbion tutorial using the zelos concrete engine. In that tutorial, the goal is to force execution of a binary down the path that prints "Executing stage 2 fake malware V2" instead of the default message:

$ ./not_packed_elf64
[+] Parsing malware configuration
[+] Virtual environment detected!

The not_packed_elf64 binary is duplicated from the angr-binaries repository. The reimplemented tutorial example script will concretely execute up to the decision point, solve for a value that will ultimately drive excution to the desired path, write that value into zelos, then resume execution in zelos. The basic workflow is to start the binary via the zelos.zdbserver, then run the angr script that utilizes the zdbserver, for instance:

Terminal 1 (zelos):

(zelos) $ python -m zelos.zdbserver not_packed_elf64

Terminal 2 (angr):

(angr) $ python3 -m angr_zelos_target.example

Terminal 2 Output:

[0] Created angr_zelos project for 'angr_zelos_target/example/not_packed_elf64'
[1] Got to decision point concretely.
[2] Symbolically finding second stage @ 0x400bb6
[3] Executing concretely until exit @ 0x65310d
[4] DONE.

Terminal 1 Output:

[main] [SYSCALL] brk ( addr=0x0 ) -> 0x900000a4
[main] [SYSCALL] openat ( dirfd=0xffffff9c, pathname=0xb229170 ("not_packed_elf64"), flags=0x80000 ) -> 18
...
...
[main] [SYSCALL] brk ( addr=0x90022000 ) -> 0x90022000
Breakpoint "bp_400af3"
[StdOut]: 'bytearray(b'[+]Parsing malware configuration\n\n[+] Executing stage 2 fake malware V2\n\n')'
[main] [SYSCALL] write ( fd=0x1, buf=0x90000310 ("[+]Parsing malware configuration [+] Executing stage 2 fake malware V2"), count=0x3a ) -> 3a
[main] [SYSCALL] exit_group ( status=0x0 ) -> void

AGPL v3