> For the complete documentation index, see [llms.txt](https://docs.vida.id/identity-stack/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.vida.id/identity-stack/verify/liveness/overview/image-manipulation-detection.md).

# Image Manipulation Detection

Image manipulation detection is an additional layer of check to identify morphed patterns commonly found in photoshopped images, swapped selfies, and deep fakes. This provides a comprehensive liveness detection solution, enhancing the overall security of the process. 

The below visualisation represents how the combined evaluation of liveness and image manipulation scores guides the system's decisions accurately.

<figure><img src="/files/c3FVVAwUOQZBoit6N9ho" alt=""><figcaption></figcaption></figure>

<figure><img src="https://app.gitbook.com/o/HFo4Mq4hPbfFvhTREpfY/s/zqikP7fR7hLesltIaRUc/product-overview/liveness-detection/image-manipulation-detection" alt=""><figcaption></figcaption></figure>

{% hint style="info" %}
Image Manipulation detection is integrated into our eKYC API, Liveness API, Liveness SDK, and eKYC Web 2.0.

It can be enabled be on requirement. Please contact the [<mark style="color:blue;">VIDA support team</mark>](mailto:support@vida.id)<mark style="color:blue;">,</mark> to get it enabled.
{% endhint %}

**Swapped Face Detection**

Combined with current liveness detection, VIDA aims to increase the chance of detecting, mitigating, and minimizing any face morph or face wrap occurrence, especially one that’s generated by deep fake technology.

<figure><img src="/files/xn3m3qOag4rVaKtkWbuQ" alt="" width="375"><figcaption></figcaption></figure>

### **How it Works**

The image manipulation feature employs an advanced algorithm to identify manipulations in selfie images, such as those created through photo editing or morphing techniques. This feature acts as a secondary layer in the backend to safeguard against deep fakes or edited selfies, with a default threshold setting at `0.95`.

<figure><img src="/files/pv4ZzyyWy4ze3g5lmJkV" alt="" width="375"><figcaption></figcaption></figure>

**Scoring & Thresholds:**

Image manipulation scores range from `0` to `1`. The threshold for determining manipulation is set at `0.95`.

* A score below `0.95` indicates a genuine user, and the selfie is accepted.
* A score of `0.95` or higher signifies a manipulated selfie, leading to rejection.

<figure><img src="/files/2goszt9B0XZZBB4OeF44" alt=""><figcaption></figcaption></figure>

<table><thead><tr><th width="486.65472412109375">Score-based use case</th><th>Decision result</th></tr></thead><tbody><tr><td>When the liveness score &#x3C; <code>0.95</code> and image manipulation score &#x3C; <code>0.95</code></td><td>Selfie is accepted</td></tr><tr><td>When the liveness score &#x3C; <code>0.95</code> and image manipulation score >= <code>0.95</code></td><td>Selfie is rejected</td></tr><tr><td>When the liveness score >= <code>0.95</code> and image manipulation score &#x3C; <code>0.95</code></td><td>Selfie is rejected</td></tr><tr><td>When the liveness score >= <code>0.95</code> and image manipulation score >= <code>0.95</code></td><td>Selfie is rejected</td></tr></tbody></table>