When executing fuzzing experiment on embedded environment one often faces challenge of performing multiple tasks in repeatable and observable manner, for example: reset board, ensure embedded software booted, send fuzzing data using selected link, monitor peripheral state to detect changes in behaviour etc.
This is what Emfuzzer helps to orchestrate: it runs various tools and scripts in specific manner, then gathers their results for further inspections.
Note: although focused on fuzzing and embedded systems, Emfuzzer can help with any software-related experiments, that require repeatable order of tasks and results capture.
Emfuzzer is available on PyPI, it is recommended to install
it in isolated environment, either by using Python venv or
tools like pipx.
python -m venv .venv
source .venv/bin/activate
pip install emfuzzerpipx install emfuzzerTo run experiments simply run:
emfuzzer --config=experiment.json test1.bin test2.binFor each specified data file steps from experiment.json
will be executed and gathered results stored in file named
emfuzzer-CURRENTDATE.json. Application will output logs
to the console and also store them in .log file next to
the .json results file. The prefix for output files can
be modified using --output-prefix command line switch.
See default-config.json in source directory for example
of experiment definition (this file can be safely used -
the "experiment" calls cat on each passed file).
To obtain complete command line switches documentation call
emfuzzer --help.
Each data file passed to the emfuzzer represents a single Test Case. For each test case following experiment steps will be performed:
- Setup tasks will be executed and their results stored.
- Monitoring tasks will be started.
- Injector will use Test Case data to perform the main experiment task (e.g. fuzzing data injection).
- Injector will observe system behaviour to capture result of the injection.
- Monitoring tasks will finish, their results stored.
- Check tasks will be executed and their results stored.
- Go to 1 for next Test Case.
Experiment configuration is stored in JSON format.
See chapters below for list of all types of injectors, tasks and monitors (with arguments).
Below is the default-config.json with comments:
Injectors are the tools that take the experiment data, use it to "inject" it into the system, and then observe the effects. Currently supported injectors' types:
subprocess- execute script and captures its exit code. Arguments:cmd- (list of strings) command to be executed, data will be passed as the last argument to the callshell- (boolean) true when shell should be used todo interpret the commandtimeout- (float) time to wait for command to finish
coap- CoAP (Constrained Application Protocol) injector data will be sent over UDP to specified address, success when positive response is received from the system. Arguments:target- dictionary containinghostandportof the targetresponse_timeout- (float) timeout to wait for CoAP response after sending the dataobservation_timeout- (float) additional time for detecting any unexpected messages after the response
Sub Tasks are tasks that for each case can be executed as setups or checks.
Note: failure of "setup" does not interrupt the test case execution - it is logged and stored in results, next steps are still executed, to be analyzed later.
Monitors are tasks that their execution is started after setups, then they are active during the injection and finish before checks.
Available tasks:
subprocess- execute script and capture its exit code. Arguments:cmd- (list of strings) command to be executedshell- (boolean) true when shell should be used todo interpret the commandfinish- configuration of finishing the task:signal- (string) signal name to be sent to the task (can beNONE)timeout- (float) time to wait for command to finish (starts after signal is sent)
ping_stable- pings a target number of times, expects all pings to reply. Arguments:host- (string) host to be checkedcount- (integer) number of pings to sentinterval- (integer) interval between pings
ping_alive- pings a target and expects first response Arguments:host- (string) host to be checkedtimeout- (float) timeout to wait for responseinterval- (integer) interval between pings
remote- executes command over SSH and captures its exit code. Arguments:connection- dictionary containing:hostportusernamepassword
command- (string) command to be executedstart_key- (string) string expected in the output of the executed command for the command to be considered "started successfully"start_timeout- (float) timeout for the start of the commandfinish- configuration of finishing the task:signal- (string) signal name to be sent to the task (can beNONE)timeout- (float) time to wait for command to finish (starts after signal is sent)
coap_monitor- listens for CoAP responses Arguments:target- dictionary containinghostandportof the targetresponse_timeout- (float) timeout to wait for CoAP response after sending any data (useful only when used as Injector)observation_timeout- (float) additional time for detecting any unexpected messages when monitoring finishes
{ "delays": { "between_cases": 0.2, // delay between Test Cases "before_inject": 1 // delay after all setups }, "injector": { "type": "subprocess", // type of injection "args": { // arguments for given injector "cmd": [ "cat" ], "shell": false, "timeout": 1 } }, "case": { // for each case "setups": [ // list of setups { "type": "subprocess", // type of setup tasks "name": "setup", // name used in results "args": { // arguments for given setup "cmd": [ "echo", "SETUP" ], "finish": { "timeout": 0.5, "signal": "NONE" }, "shell": false } }, { // second setup "type": "ping_alive", "name": "ping", "args": { "host": "127.0.0.1", "timeout": 10, "interval": 1 } } ], "monitoring": [], // monitoring tasks "checks": [ // list of checks tasks { // same as setups "type": "ping_stable", "name": "ping", "args": { "host": "127.0.0.1", "count": 2, "interval": 1 } }, { "type": "subprocess", "name": "teardown", "args": { "cmd": [ "echo", "TEARDOWN" ], "finish": { "timeout": 0.5, "signal": "NONE" }, "shell": false } } ] } }