Modes of Operation

Frida provides dynamic instrumentation through its powerful instrumentation core Gum, which is written in C. Because such instrumentation logic is prone to change, you usually want to write it in a scripting language so you get a short feedback loop while developing and maintaining it. This is where GumJS comes into play. With just a few lines of C you can run a piece of JavaScript inside a runtime that has full access to Gum’s APIs, allowing you to hook functions, enumerate loaded libraries, their imported and exported functions, read and write memory, scan memory for patterns, etc.

Table of contents

  1. Injected
  2. Embedded
  3. Preloaded


Most of the time, however, you want to spawn an existing program, attach to a running program, or hijack one as it’s being spawned, and then run your instrumentation logic inside of it. As this is such a common way to use Frida, it is what most of our documentation focuses on. This functionality is provided by frida-core, which acts as a logistics layer that packages up GumJS into a shared library that it injects into existing software, and provides a two-way communication channel for talking to your scripts, if needed, and later unload them. Beside this core functionality, frida-core also lets you enumerate installed apps, running processes, and connected devices. The connected devices are typically iOS and Android devices where frida-server is running. That component is essentially just a daemon that exposes frida-core over TCP, on localhost:27042.


It is sometimes not possible to use Frida in Injected mode, for example on jailed iOS and Android systems. For such cases we provide you with frida-gadget, a shared library that you’re supposed to embed inside the app that you want to instrument. This library starts running as soon as the dynamic linker executes its constructor function, and exposes the same interface as frida-server does, listening on localhost:27042. The only difference is that the lists of running processes and installed apps only contain a single entry, which is for the app itself. The process name is always just Gadget, and the installed app’s identifier is always re.frida.Gadget. In order to achieve early instrumentation we let the aforementioned constructor function block until you either attach() to the process, or call resume() after going through the usual spawn() -> attach() -> …apply instrumentation… steps. This means that existing CLI tools like frida-trace work the same ways you’re already using them.


Perhaps you’re familiar with LD_PRELOAD, or DYLD_INSERT_LIBRARIES? Wouldn’t it be cool if there was JS_PRELOAD? This is where frida-gadget, the shared library discussed in the Embedded section, also provides a second mode of operation which doesn’t involve any TCP or outside communication. All you need to do is to set the FRIDA_GADGET_SCRIPT environment variable to the path to the file containing your JavaScript.

For example on GNU/Linux, just create the file hook.js with the contents:

'use strict';

rpc.exports = {
  init: function () {
    Interceptor.attach(Module.findExportByName(null, 'open'), {
      onEnter: function (args) {
        var path = Memory.readUtf8String(args[0]);
        console.log('open("' + path + '")');

The latest frida-gadget for your OS can be found on GitHub.

Now just set two environment variables and launch your target process:

LD_PRELOAD=/path/to/ \
FRIDA_GADGET_SCRIPT=/path/to/hook.js \
cat /etc/hosts

Use DYLD_INSERT_LIBRARIES on macOS and iOS. Note that /bin/cat won’t work on El Capitan and above, as it ignores such attempts for system binaries.

You may also add FRIDA_GADGET_ENV=development while developing your instrumentation logic, which will make frida-gadget watch your file for changes and automatically reload the script whenever it changes on disk. This will even work if your script hooks functions, like in this example above, as all hooks are reverted automatically when the old version of the script is unloaded.

The reason we expose an init() method using Frida’s RPC feature is because frida-gadget will call it and wait for it to return until it lets the program continue executing its entrypoint. This means you can return a Promise if you need to do something asynchronous, e.g. Memory.scan() to locate a function you want to instrument, and guarantees that you won’t miss any early calls. You may also expose a dispose() method if you need to perform some explicit cleanup when the process exits or your script get unloaded before the new version is loaded from disk (which happens with FRIDA_GADGET_ENV=development).

For debugging you can use console.log(), console.warn(), and console.error(), which will print to stdout/stderr.