Tamper-resilient storage & keying

**Description**
Implement a secure, modular storage layer that focuses on **resilience against tampering or post-compromise forensics**. This is not about detecting tamper events, but ensuring that once a trigger fires, **data becomes cryptographically irrecoverable**.

This storage layer should tie into existing trigger mechanisms to wipe RAM-resident keys, making data unreadable even with physical access working alongside tamper detection layers (see #2) for full lifecycle defense.

**Components**

* Ephemeral keying for encrypted storage

* RAM-resident encryption keys (non-persistent)
* Keys wiped on trigger (e.g., keyboard phrase, USB event, network packet)
* Integration with dm-crypt / LUKS volumes

* One-time pad (OTP) or XOR-based multi-layer key splits

* Distribute decryption key into multiple parts
* Require all pieces for full recovery
* Optional: store parts across different locations (e.g., device + remote peer)

**Goals**

* Use standard Linux mechanisms (e.g., dm-crypt, LUKS) where possible
* Abstract encryption/key-wiping logic behind an internal API
* Ensure key material never touches disk or swap
* Optionally: allow key splitting via simple XOR or Shamir's Secret Sharing

**Optional extensions**

* TPM-backed key storage with time-based unlock windows
* Remote key escrow with secure request protocol
* Support for volatile in-memory filesystems (e.g., `tmpfs` + encrypted overlay)

El issue describe la integración con mecanismos de activación existentes (ver issue #2) y módulos del kernel de Linux como dm-crypt/LUKS. Comience examinando el código fuente del módulo del kernel en el repositorio para comprender el sistema de activación. Busque código existente de almacenamiento o cifrado. El objetivo es diseñar una capa de almacenamiento seguro que borre las claves residentes en RAM al activarse, haciendo que los datos sean irrecuperables. 'Hecho' significa una API modular que se conecte a los activadores y utilice mecanismos de cifrado estándar de Linux.

Escrito por el modelo de indexación a partir del texto del issue.