TL;DR
This paper introduces the Secure Transport (ST) facility within the DREMS operating system, designed for distributed spacecraft. ST provides a robust network transport layer that enforces strict information flow control and isolation between applications from different security domains, crucial for shared, multi-organizational space missions.
What Problem Does This Paper Address?
Modern space missions are increasingly moving towards distributed architectures, where multiple independent satellites (forming a "fractionated spacecraft") work together. This approach offers cost savings and increased adaptability compared to traditional monolithic satellites. However, it introduces significant challenges, particularly in security. When these distributed systems host applications from various organizations with different security requirements and policies, ensuring data isolation and preventing unauthorized access becomes paramount. The platform needs to manage and compartmentalize information flows securely, even when applications are not fully trusted. This paper tackles the problem of building a secure, resilient, and adaptable software platform for such distributed spacecraft that can handle mixed-criticality applications and data from different security domains.
Methods
The core contribution is the Secure Transport (ST) mechanism, integrated into the DREMS (Distributed Realtime Managed System) operating system. ST acts as a network transport layer that enforces information flow partitions based on Multi-Level Security (MLS) labels. These labels represent security classifications and need-to-know categories, dictating how information can flow between processes.
Key aspects of the ST approach include:
- Mandatory Access Control (MAC) Policies: ST enforces MAC policies, ensuring that information only flows according to predefined security rules. This includes "read down" (higher security label receiving from lower) and "write up" (lower security label sending to higher) capabilities.
- Multi-Domain Labels: A novel aspect is the support for "multi-domain labels" to facilitate secure communication among actors from different organizations, going beyond standard MLS.
- Endpoints: Communication is managed through endpoints, analogous to socket handles. These are categorized into Loopback (LBE), Local Message (LME), Remote Message (RME), and Kernel Message (KME) endpoints, each with specific restrictions.
- Flows: Communication between endpoints is governed by "flows," which are explicit configurations defining allowed send/receive relationships. Both source and destination flows must be mutually compatible for communication to occur.
- Trusted Computing Base (TCB): The ST infrastructure is part of the system's TCB, meaning it's a trusted component responsible for enforcing security policies. Configuration of endpoints and flows is restricted to privileged tasks within the TCB.
- Quality of Service (QoS) and Cryptography: ST supports various QoS properties (e.g., protocol selection like UDP/SCTP) and aims to provide cryptographically secure communications (confidentiality, integrity, authenticity) through mechanisms like IPSec, though IPSec support is noted as a work in progress.
Datasets
Not discussed in the paper. The evaluation uses a custom experimental setup with two computing nodes.
Key Insights and Ideas
- Security as a Foundation, Not an Afterthought: The paper strongly emphasizes that security, particularly information flow control and isolation, must be a fundamental design principle for distributed spacecraft, not an add-on.
- The Power of MLS and Multi-Domain Labels: Leveraging MLS labels is a standard approach, but extending it to "multi-domain" labels is crucial for enabling secure collaboration between independent organizations on the same platform. This allows for compartmentalization even when the overall system is shared.
- Explicit Flow Control is Essential: Relying solely on network protocols is insufficient. The concept of "flows" provides a fine-grained, policy-driven mechanism to explicitly define and enforce communication paths, preventing unintended data leakage.
- Actor-Centric Communication: The use of "actors" with globally unique identifiers simplifies communication management, allowing actors to be migrated between nodes without reconfiguring communication paths.
- Balancing Security and Performance: The implementation details, particularly the use of read/write mutexes and spinlocks, highlight the careful engineering required to balance strict security enforcement with the need for efficient inter-process communication, especially in real-time systems.
Key Results
The paper presents a conceptual design and implementation of the Secure Transport facility. An empirical evaluation using a simple sender-receiver setup demonstrated approximately 93% utilization of UDP for bidirectional communication, suggesting the ST facility does not impose an excessive performance overhead for basic data transfer.
Limitations and Future Work
- Centralized Label Model: The current implementation uses a centralized label model. The authors plan to investigate extensions to support a fully decentralized label model in the future.
- IPSec and Encrypted Physical Layer: IPSec support in Secure Transport is noted as a work in progress. The implementation of an encrypted physical medium is explicitly out of scope for the current ST implementation.
- Experimental Validation of Hypothesis: The paper mentions a working hypothesis regarding the scheduling of reactor threads for message delivery and notes that this has yet to be experimentally validated.
- Multi-Domain Label Implementation: While multi-domain labels are discussed as a concept, the paper focuses on the core ST mechanism and its MLS foundation. The full implications and implementation details of multi-domain labels for inter-organizational communication would be a natural area for further exploration.
Summary updated