Awesome
FA
What is it?
FA stands for Firmware Analysis and intended For Humans.
FA allows one to easily perform code exploration, symbol searching and other functionality with ease.
Usually such tasks would require you to understand complicated APIs, write machine-dependant code and perform other tedious tasks. FA is meant to replace the steps one usually performs like a robot (find X string, goto xref, find the next call function, ...) in a much friendlier and maintainable manner.
The current codebase is very IDA-plugin-oriented. In the future I'll consider adding compatibility for other disassemblers such as: Ghidra, Radare and etc...
Pull Requests are of course more than welcome :smirk:.
Requirements
Supported IDA 8.0+.
In your IDA's python directory, run:
python -m pip install -r requirements.txt
And for testing:
python -m pip install -r requirements_testing.txt
Where to start?
Before using, you should understand the terminology for: Projects, SIG files and Loaders.
So, grab a cup of coffee, listen to some nice music, and please devote a few minutes of your time into reading this README.
Projects
The "project" is kind of the namespace for different signatures. For example, either: linux, linux_x86, linux_arm etc... are good project names that can be specified if you are working on either platforms.
By dividing the signatures into such projects, Windows symbols for example won't be searched for Linux projects, which will result in a better directory organization layout, better performance and less rate for false-positives.
The signatures are located by default in the signatures
directory.
If you wish to use a different location, you may create config.ini
at FA's root with the following contents:
[global]
signatures_root = /a/b/c
SIG format
The SIG format is a core feature of FA regarding symbol searching. Each SIG file is residing within the project directory and is automatically searched when requested to generate the project's symbol list.
The format is Hjson-based and is used to describe what you, as a human, would do in order to perform the given task (symbol searching or binary exploration).
SIG syntax (single):
{
name: name
instructions : [
# Available commands are listed below
command1
command2
]
}
Each line in the instructions
list behaves like a shell
command-line that gets the previous results as the input
and outputs the next results
to the next line.
Confused? That's alright :grinning:. Just look at the examples below
User may also use his own python script files to perform
the search. Just create a new .py
file in your project
directory and implement the run(interpreter)
method.
Also, the project's path is appended to python's sys.path
so you may import your scripts from one another.
To view the list of available commands, view the list below
Examples
Finding a global struct
{
name: g_awsome_global,
instructions: [
# find the byte sequence '11 22 33 44'
find-bytes '11 22 33 44'
# advance offset by 20
offset 20
# verify the current bytes are 'aa bb cc dd'
verify-bytes 'aa bb cc dd'
# go back by 20 bytes offset
offset -20
# set global name
set-name g_awsome_global
]
}
Find function by reference to string
{
name: free
instructions: [
# search the string "free"
find-str 'free' --null-terminated
# goto xref
xref
# goto function's prolog
function-start
# reduce to the singletone with most xrefs to
max-xrefs
# set name and type
set-name free
set-type 'void free(void *block)'
]
}
Performing code exploration
{
name: arm-explorer
instructions: [
# search for some potential function prologs
arm-find-all 'push {r4, lr}'
arm-find-all 'push {r4, r5, lr}'
thumb-find-all 'push {r4, lr}'
thumb-find-all 'push {r4, r5, lr}'
# convert into functions
make-function
]
}
Performing string exploration
{
name: arm-string-explorer
instructions: [
# goto printf
locate printf
# iterate every xref
xref
# and for each, go word-word backwards
add-offset-range 0 -40 -4
# if ldr to r0
verify-operand ldr --op0 r0
# go to the global string
goto-ref --data
# and make it literal
make-literal
]
}
Finding enums and constants
{
name: consts-finder
instructions: [
# goto printf
locate printf
# iterate all its function lines
function-lines
# save this result
store printf-lines
# look for: li r7, ???
verify-operand li --op0 7
# extract second operand
operand 1
# define the constant
set-const IMPORTANT_OFFSET
# load previous results
load printf-lines
# look for: li r7, ???
verify-operand li --op0 8
# get second operand
operand 1
# set this enum value
set-enum important_enum_t some_enum_key
]
}
Adding struct member offsets
{
name: structs-finder
instructions: [
# add hard-coded '0' into resultset
add 0
# add a first member at offset 0
set-struct-member struct_t member_at_0 'unsigned int'
# advance offset by 4
offset 4
# add a second member
set-struct-member struct_t member_at_4 'const char *'
# goto function printf
locate printf
# iterate its function lines
function-lines
# look for the specific mov opcode (MOV R8, ???)
verify-operand mov --op0 8
# extract the offset
operand 1
# define this offset into the struct
set-struct-member struct_t member_at_r8_offset 'const char *'
]
}
Finding several functions in a row
{
name: cool_functions
instructions: [
# find string
find-str 'init_stuff' --null-terminated
# goto to xref
xref
# goto function start
function-start
# verify only one single result
unique
# iterating every 4-byte opcode
add-offset-range 0 80 4
# if mnemonic is bl
verify-operand bl
# sort results
sort
# store resultset in 'BLs'
store BLs
# set first bl to malloc function
single 0
goto-ref --code
set-name malloc
set-type 'void *malloc(unsigned int size)'
# go back to the results from 4 commands ago
# (the sort results)
load BLs
# rename next symbol :)
single 1
goto-ref --code
set-name free
set-type 'void free(void *block)'
]
}
Conditional branches
{
name: set_opcode_const
instructions: [
# goto printf function
locate printf
# goto 'case_opcode_bl' if current opcode is bl
if 'verify-operand bl' case_opcode_bl
# make: #define is_bl (0)
clear
add 0
set-const is_bl
# finish script by jumping to end
b end
# mark as 'case_opcode_bl' label
label case_opcode_bl
# make: #define is_bl (1)
clear
add 1
set-const is_bl
# mark script end
label end
]
}
Python script to find a list of symbols
from fa.commands.find_str import find_str
from fa.commands.set_name import set_name
from fa import context
def run(interpreter):
# throw an exception if not running within ida context
context.verify_ida('script-name')
# locate the global string
set_name(find_str('hello world', null_terminated=True),
'g_hello_world', interpreter)
Python script to automate SIG files interpreter
TEMPLATE = '''
find-str '{unique_string}'
xref
function-start
unique
set-name '{function_name}'
'''
def run(interpreter):
for function_name in ['func1', 'func2', 'func3']:
instructions = TEMPLATE.format(unique_string=function_name,
function_name=function_name).split('\n')
interpreter.find_from_instructions_list(instructions)
Python script to dynamically add structs
from fa.commands.set_type import set_type
from fa import fa_types
TEMPLATE = '''
find-str '{unique_string}'
xref
'''
def run(interpreter):
fa_types.add_const('CONST7', 7)
fa_types.add_const('CONST8', 8)
foo_e = fa_types.FaEnum('foo_e')
foo_e.add_value('val2', 2)
foo_e.add_value('val1', 1)
foo_e.update_idb()
special_struct_t = fa_types.FaStruct('special_struct_t')
special_struct_t.add_field('member1', 'const char *')
special_struct_t.add_field('member2', 'const char *', offset=0x20)
special_struct_t.update_idb()
for function_name in ['unique_magic1', 'unique_magic2']:
instructions = TEMPLATE.format(unique_string=function_name,
function_name=function_name).split('\n')
results = interpreter.find_from_instructions_list(instructions)
for ea in results:
# the set_type can receive either a string, FaStruct
# or FaEnum :-)
set_type(ea, special_struct_t)
Aliases
Each command can be "alias"ed using the file
found in fa/commands/alias
or in <project_root>/alias
Syntax for each line is as follows: alias_command = command
For example:
ppc32-verify = keystone-verify-opcodes --bele KS_ARCH_PPC KS_MODE_PPC32
Project aliases have higher priority.
Loaders
Loaders are the entry point into running FA. In the future we'll possibly add Ghidra and other tools.
Please first install the package as follows:
Clone the repository and install locally:
# clone
git clone git@github.com:doronz88/fa.git
cd fa
# install
python -m pip install -e .
IDA
Within IDA Python run:
from fa import ida_plugin
ida_plugin.install()
You should get a nice prompt inside the output window welcoming you into using FA. Also, a quick usage guide will also be printed so you don't have to memorize everything.
Also, an additional FA Toolbar
will be added with quick functions that
are also available under the newly created FA
menu.
A QuickStart Tip:
Ctrl+6
to select your project, then Ctrl+7
to find all its defined symbols.
You can also run IDA in script mode just to extract symbols using:
ida -S"fa/ida_plugin.py <signatures-root> --project-name <project-name> --symbols-file=/tmp/symbols.txt" foo.idb
ELF
In order to use FA on a RAW ELF file, simply use the following command-line:
python elf_loader.py <elf-file> <signatures_root> <project>
Available commands
See commands.md