In-App virtualization is becoming a popular subject on mobile platform. This unique feature of Android and alike allows developers to build hundred of virtualization capable apps, with millions of users, to allegedly enhance privacy or offer muli-spaces. When considering future of malwares, it is generally admitted that In-app virtualization may weaken security of (1) the host app and (2) the virtualized apps among themselves. Recent studies also show that host app with malicious intent are a danger to virtualized apps. However, we believe that the danger of this technique is far greater for the whole android ecosystem as a non-maliciously payloaded (no requested permissions, no keylogging hooks, for instance) host and an even non-root device can attempt to deceive any app installed on the device. We will show in this talk how trivial it is to build a malware platform that rely on few if no permission yet is capable of launching advanced, targeted and undetected attacks. Such attacks include stealing of user personal data, stealing of user on-line identity, eavesdropping of network connection, even if SSL secured, and live in-memory patching of the ART structures to redirect or proxify core Android methods.