A secure IoT gateway platform.

The IDS Trusted Connector is an open IoT edge gateway platform and an implementation of the Trusted Connector in the Industrial Data Space Reference Architecture, following the DIN Spec 27070 and ISO 62443-3 standards.

Use the Trusted Connector to connect sensors with cloud services and other Connectors, using a vast range of protocol adapters. Security mechanisms like secure boot, remote platform attestation and data usage control allow you stay in control over your data. The Trusted Connector is open source and built on open standards to avoid vendor lock-in.

Main Features

App Isolation

Apps running in the Trusted Container are isolated from each other and from the Internet. This way, apps cannot leak data or harm each other or the Connector.

Cross-Enterprise Identity Management

Trusted Connectors authorize each other using a dynamic attribute provisioning service (DAPS). This allows easy cross-enterprise authorization without the hassle of X509v3 cross certification.

Trusted Platform

With the trust|me container management and secure boot, the Trusted Connector makes use of a TPM to gurantee the integrity of the software stack from bare metal to applications. A remote attestation protocol makes sure you can rely on your remote party's integrity.

Data Usage Control

Control the flow of data with LUCON policies and bind remote data usage to obligations. Rely on attested trustworthy platforms to enforce obligations and make sure your message routes comply with security policies and legislation.

App Isolation

The isolation between Apps is based on the isolation between containers. The Core Container is a privileged container running the Core Container Stack, which takes care of all management and routing tasks. The architecture provides OS-level virtualization. This means the kernel is shared between all container instances. Kernel security mechanisms are used to provide user space virtualization. Access to hardware interfaces (like network interfaces) can be limited to specific containers. In a default configuration, only the core container is allowed to communicate with the outside. Service containers can only be accessed over data routes configured there.

Isolation works based on kernel features: Each container runs in a separate namespace. Thus, namespaces virtualize the kernel resources. This makes containers unaware of components in a different namespace. Cgroups allow for flexible allocation and enforcement of process group policies. This way, process groups can be formed to create a container. This also allows one to define policies regarding CPU and memory limitations as well as other hardware resources, as external interfaces. The Linux Security Module (LSM) enforces protection policies by introducing hooks at critical points of execution. Capabilities are used to assign processes with a fine-grained set of permissions to regulate access to kernel objects. Finally, Full Disk Encryption (FDE) is used to encrypt all persisted container storage. This approach is compatible with Linux Containers (LXC).

Trusted Platform

The Trusted Connector supports two container management layers: Docker and trustme.

Docker (container engine) logo
  • For standard Linux stacks, leveraging the powerful Docker ecosystem
  • Support for x86 and Raspberry Pi (ARM) platforms
  • For highly secure embedded platforms
  • Support for remote attestation & platform integrity measurement of the trusted platform
  • Support for ARM, x86, PowerPC architectures

No matter whether Docker or trust|me is used, any image from the Trusted Connector App Store or from any Docker registry can be loaded into the Trusted Connector and launched as an application. With trust|me, Docker images must be converted into a trust|me guestOS format before installation, using the automated converter tool.

The Trusted Connector features the secure container management layer trust|me as an alternative to Docker. Though its basic mechanisms are similar to Docker (namespaces, cgroups and chroot), trust|me was developed as a security architecture including secure boot, platform integrity measurements, and a hardened kernel.

For mandatory access control, trust|me adds a custom Linux Security Module (LSM) to the Linux kernel, which can be combined with SELinux using LSM stacking. In addition, the trust|me architecture includes a secure element for tamper-proof key storage and a TPM 2.0 interface for (remote) attestation of the platform integrity using the IDSCP protocol.

For a more detailed description of the architecture see our publication Manuel Huber, Julian Horsch, Michael Velten, Michael Weiss, Sascha Wessel. A Secure Architecture for Operating System-Level Virtualization on Mobile Devices. In Information Security and Cryptology 2016, pp. 430-450

Core Container

The Core Container provides the main functionality of the Trusted Connector and is the sole intermediate between application containers and the Internet. It runs on top of the container management layer and hosts different services for managing applications, setting up data routes and connecting to other Trusted Connectors. Based on Spring Boot, it provides a REST API. The main services are

  • Usage Control: Data flows and the usage of messages can be controlled using the LUCON policy framework.
  • Route Manager: Message routing based on Apache Camel. The basic installation contains protocol adapters for
    • IDSCP2: A secure TCP/IP-based protocol between Trusted Connectors, including remote attestation functionalities
    • OPC-UA (based on Eclipse Milo)
    • MQTT (based on Eclipse Paho)
    • JMS
    • REST
    • HTTP(S)
    • Websockets

    Apache Camel supports more than 200 protocol adapters which can be used by the Trusted Connector.

  • IDS Protocol: The IDSC protocol establishes trust and sets up a secure messaging channel between Trusted Connectors. Integrity of platform stacks are measured and remotely attested, trust levels and usage control policies are negotiated between endpoints. The IDSC protocol is available as an endpoint in Apache Camel routes.
  • Management GUI: An administration web interface

Application containers

Applications come in the form of Docker containers. No matter whether Docker or trustme is used, any image from a Docker registry can be pulled into the Trusted Connector and launched as an application. Without further configuration, application containers are initially isolated from each other and restricted in a virtual network with the Core Platform which blocks outbound traffic. That is, even malicious applications cannot interfere with the running system. This is a major advantage over other, purely OSGi-based application frameworks.

Cross-Enterprise Identity Management

To connect data sources across trust boundaries, a homogenous concept for identity management is mandatory. All IDS Connectors possess a unique identifier embedded in a X.509 certificate that identifies every connector instance. Attributes that can be dynamic in nature (e.g., IDS membership or certification status) are embedded into a dynamic token. This way, a flexible and extensible identity management is achieved.

Identity tokens

Every connector needs three identity tokens:

  • A device certificate (X.509v3)
  • A TLS connection certificate (X.509v3)
  • A ‘Dynamic Attribute Token’ (OAuth Access Token)

Every connector needs a certificate issued by the Device-CA. This certificate serves as the root of identity. The contents of this certificate are kept at minimum to avoid the need for later revocation in case of changing attributes. This certificate needs to be manually deployed during connector setup.

The TLS connection certificate is used for TLS tunneling. This certificate is automatically requested by the connector by interacting with a ACME server that is integrated into the TLS Sub-CA. So no manual deployment is needed.

The ‘Dynamic Attribute Token’ is an OAuth Access Token, signed by the Dynamic Attribute Provisioning Service (DAPS). This is a short-lived token that contains attributes that the connector possesses.

Authorization workflow

When a resource on a connector is accessed, an access token needs to be presented by the requesting connector. This is performed automatically with a run of the IDSCP.

The general workflow is:

  • A requesting connector presents its Device Certificate to the DAPS to receive a Dynamic Attribute Token (DAT).
  • This DAT can be used to directly access a resource provided by another Connector. An alternative configuration option would be to introduce a local or use-case specific Authorization Service that provides a custom access token. This is completely optional and thus marked grey.
  • The respective token is then handed to a Connector with every resource access request.

Dynamic Access Token issuance

  • A: call C3/token endpoint with Client Credentials (X.509 Cert)
  • B: Issue JWT-1{attribute_list, client_id, aud: idsAS:*}
  • C (optional): Hand in JWT-1, request JWT-2 {scope: C1/PS}
  • D: Use Rule Engine for access decision, issue JWT-2{aud: C1}
  • E: Connect via IDSCP (Auth by JWT-2)

An example for the attribute list would be certification status or membership in the IDS.

Data Usage Control

A central part of the data exchange between connectors is the ability to document and control data usage. In addition to access control, Usage Control influences how data may be processed throughout its lifetime. To enable Usage Control, we created LUCON, a policy framework for ‘Logic Based Usage Control’. This frameworks allows to control data flows between Apps and Connector instances. Data is labeled by LUCON as soon as it is processed by the Connector. While the data is processed by a data route, these labels are transported along with the data and, depending on the defined policies, are removed, transformed or extended. Usage of this data can be limited or invalid data flows can be suppressed. It is impossible to associate obligations to data that allow to limit the way data can be accessed or to enforce specific actions to be carried out.

Data that is marked private can be hindered from leaving a Connector. It is also possible to enforce data undergoing anonymization provided by a dedicated service before being accessed from the outside. An example for an obligation is a logging event were every access to a data item leads to an entry in the audit log. The policies are transmitted along with the data with the IDS Communication Protocol (IDSCP).

For more information, see the the dedicated section on Usage Control