> For the complete documentation index, see [llms.txt](https://docs.talsec.app/appsec-articles/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.talsec.app/appsec-articles/articles/medusa-frida-with-the-work-already-done-and-why-hook-detection-is-the-answer.md).

# Medusa: Frida with the work already done, and why hook detection is the answer

Most writing about attack tooling reaches for novelty. This article does the opposite, because the accurate description of **Medusa** is more useful than an exciting one.

Medusa is a modular framework built on top of **Frida**. It does not introduce a new way to instrument a running application. It packages the ways that already existed into reusable modules, so that tasks which previously required writing a Frida script now require selecting one.

That is a real change, and it is worth understanding. But it is a change in *who can do this*, not in *what can be done*. For defenders, that distinction determines the entire response.

<figure><img src="https://1548930415-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FNjTFXsqCLQ3RU2oA2uHC%2Fuploads%2Fe61UP3oNV0PeehGDwa6j%2FMedusa%20Frida%20with%20the%20work%20already%20done%2C%20and%20why%20hook%20detection%20is%20the%20answer.png?alt=media&amp;token=9adb8cca-ba45-4d9b-91af-f95bdda0c4da" alt=""><figcaption></figcaption></figure>

## What is Frida &#x20;

[Frida](https://frida.re/) is a dynamic instrumentation toolkit. It attaches to a running process and lets an operator inspect and modify its behaviour at runtime — reading and rewriting function arguments and return values, tracing calls, and altering control flow, on both Android and iOS. It is a legitimate and widely used tool in security research, QA and malware analysis.

It is also the single most common instrument in mobile application attacks, for the same reason it is good at research: it gives complete visibility and control over a process on a device theoperator owns.

{% hint style="info" icon="book-blank" %}
More about hooking [**here**](https://docs.talsec.app/appsec-articles/articles/hook-hack-defend-fridas-impact-on-mobile-security-and-how-to-fight-back).
{% endhint %}

## What Medusa adds

Using Frida directly requires knowing what you want to hook and being able to write the script that hooks it. That is a meaningful barrier - not a large one for a professional, but large enough to keep casual operators out.

Medusa lowers it. As a module framework, it provides prepared instrumentation for recurring tasks, which the operator composes rather than authors. The categories are the familiar ones from any mobile assessment: observing cryptographic operations, watching filesystem and network activity, inspecting how an app handles credentials and storage, and interacting with the app's own protections.

The effect is the same one described in [our article about Lucky Patcher](https://app.clickup.com/t/869eaymch), in a more technical domain: capability that used to require expertise becomes capability that requires selection. Your threat model should account for a larger population of operators, not for a more powerful technique.

## Why standard defenses miss it

Because it *is* Frida underneath, every reason Frida evades conventional defence applies unchanged:

* **Your APK is untouched**, so signature verification passes and static analysis finds nothing.
* **The attack exists only at runtime**, in process memory, after your app has started.
* **Your own client-side checks are in scope** - a security check inside the instrumented process can be instrumented too. This is the point that matters most and the one teams most often miss.
* **Distribution is irrelevant.** The app came from the Play Store, unmodified.

And the environmental precondition is the familiar one: instrumentation of a release build on a stock device is awkward, so this usually accompanies the root and hooking environment.

{% hint style="info" icon="book-blank" %}
We've covered root and hooking environment in following articles:

* [Zygisk and Shamiko: How modern root hiding works, and how Talsec detects it](https://docs.talsec.app/appsec-articles/articles/zygisk-and-shamiko-how-modern-root-hiding-works-and-how-talsec-detects-it.md)
* [KernelSU: kernel-level root, and why Talsec still detects it](https://docs.talsec.app/appsec-articles/articles/kernelsu-kernel-level-root-and-why-talsec-still-detects-it.md)
* [LSPosed and Xposed: how runtime hooking rewrites your app's reality, and how Talsec detects it](https://docs.talsec.app/appsec-articles/articles/lsposed-and-xposed-how-runtime-hooking-rewrites-your-apps-reality-and-how-talsec-detects-it.md)
  {% endhint %}

## What Talsec detects

The defensive answer does not change with the frontend, which is the reassuring part of thisarticle: you do not need a Medusa-specific detection, you need instrumentation detection.

### freeRASP

[**freeRASP**](https://docs.talsec.app/freerasp) reports runtime instrumentation through its `hooks` threat callback, targeting the Frida layer itself rather than any framework sitting on top of it.&#x20;

It also reports:&#x20;

* `privilegedAccess`
* `appIntegrity`
* debugger attachment
* emulator state

The correlated conditions that make instrumentation practical in the first place.

{% hint style="success" icon="shield-quartered" %}
freeRASP is free and available for [Android, iOS, Flutter, React Native, Capacitor and more](https://docs.talsec.app/freerasp/integration).&#x20;
{% endhint %}

### RASP+

[**RASP+**](https://docs.talsec.app/premium-products/product/rasp+) extends this where instrumentation is an active fraud vector: broader and actively maintained hook-surface coverage, configurable response policy rather than a raw callback, and Portal telemetry so you can see instrumentation prevalence across your fleet rather than inferring it.

{% hint style="success" icon="shield-quartered" %}
See [RASP+](https://docs.talsec.app/premium-products/product/rasp+) and the [tier comparison](https://docs.talsec.app/freerasp/features-and-pricing-plans/the-key-differences-freerasp-vs.-rasp+).&#x20;
{% endhint %}

### AppiCrypt

[**AppiCrypt**](https://docs.talsec.app/premium-products/product/appicrypt) is the layer that holds when the client is fully instrumented. If an operator can modify your process, they can eventually reach your in-app verdict, so the decisions that matter financially should be authorised by yourbackend against an integrity attestation, not by a boolean computed inside a process the attackercontrols.

## Edge cases and caveats

**Hook detection is adversarial and probabilistic.** \
Frida's ecosystem includes active work on evading detection, and detection is a maintained capability rather than a fixed guarantee. Anyone promising permanent immunity is selling something.

**Legitimate use is widespread.** \
Security researchers, QA engineers and malware analysts use Frida daily and have no interest in your app. Detection is a risk input; it is not proof of hostile intent.

**Response should be graduated.**\
Instrumentation detected during browsing is a different decision from instrumentation detected at a transfer. Immediate hard termination also gives an operator a precise signal about where your check lives.

**Do not build a Medusa-specific detection.**\
The framework layer is the wrong target. It is replaceable, and there are several alternatives in the same space. Detect the instrumentation.

**This article deliberately omits detection internals.** \
Publishing the specifics of how hook detection works helps operators more than it helps defenders.

## Security todo list

{% stepper %}
{% step %}

### **Assume your client-side security checks are observable and modifiable**

List them, and mark which financially consequential decisions currently depend on them.
{% endstep %}

{% step %}

### **Instrument the `hooks` callback**

with [freeRASP](https://docs.talsec.app/freerasp/integration) and log it alongside `privilegedAccess`. The combination is the informative signal.
{% endstep %}

{% step %}

### **Re-anchor the consequential decisions server-side with** [AppiCrypt](https://docs.talsec.app/premium-products/product/appicrypt) &#x20;

{% endstep %}

{% step %}

### **Design graduated responses**

rather than a kill switch.
{% endstep %}

{% step %}

### **Read the** [**Frida article**](https://docs.talsec.app/appsec-articles/articles/hook-hack-defend-fridas-impact-on-mobile-security-and-how-to-fight-back) **if you have not**

It is the foundation this one sits on.
{% endstep %}
{% endstepper %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.talsec.app/appsec-articles/articles/medusa-frida-with-the-work-already-done-and-why-hook-detection-is-the-answer.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
