Security Considerations for Computing-Aware Traffic Steering

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶

This Internet-Draft will expire on 26 October 2025.¶

Copyright Notice

Copyright (c) 2025 IETF Trust and the persons identified as the document authors. All rights reserved.¶

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶

Table of Contents

1. Introduction

The CATS framework is an ingress-based overlay framework for the selection of the suitable service instance(s) from a set of instance candidates. By taking into account both networking and computing metrics, the CATS framework achieve a global of dispatching service demands over the various and available edge computing resources. However, ubiquitous distributed computing resources in CATS also pose challenges to security protection. The operators of CATS may not have complete control over the nodes and therefore guarantee the security and credibility of the computing nodes themselves. Moreover, there are great differences in the security capabilities provided by computing nodes in the network, which greatly improves the breadth and difficulty of security protection.¶

This document describes various threats and security concerns related to CATS and existing approaches to solve these threats.¶

2. Terminology

This document makes use of the following terms:¶

Computing-Aware Traffic Steering (CATS): A traffic engineering approach [RFC9522] that takes into account the dynamic nature of computing resources and network state to optimize service-specific traffic forwarding towards a given service instance. Various relevant metrics may be used to enforce such computing-aware traffic steering policies. [I-D.ldbc-cats-framework]¶

CATS Service ID (CS-ID): An identifier representing a service, which the clients use to access it.¶

Service: An offering provided by a service provider and which is delivered using one or more service functions [RFC7665].¶

CATS Service Metric Agent (C-SMA): An agent that is responsible for collecting service capabilities and status, and for reporting them to a CATS Path Selector (C-PS).¶

Service request: The request for a specific service instance.¶

3. Security Issues of The Computing Resources

The ubiquitous and flexible characterictics of computing resources and the frequent connections to the computing resources will lead to the following risks:¶

• Unauthorized Access and Control¶

  Attackers may exploit vulnerabilities in interfaces or APIs to gain unauthorized access, potentially hijacking computational resources or manipulating task execution.¶

• Data Confidentiality Breaches¶

  Sensitive data processed by computing resources (e.g., model parameters in ML workloads) could be intercepted during transmission or compromised through insecure memory handling.¶

• Denial-of-Service (DoS) Threats¶

  Malicious actors may flood computing resources with forged computation requests, degrading service availability or disrupting task scheduling.¶

To address these risks, CATS implementations COULD adopt the following safeguards:¶

• Secure Communication Frameworks¶

  • TLS 1.3 [RFC8446] could be adopted for all control-plane and data-plane communications.¶

  • Certificate-based mutual authentication could be implemented using IETF SUIT [RFC9019] for Computing Service to C-SMA interactions.¶

• Granular Access Control¶

  • Role-based access policies (RBAC) aligned with AAA architecture could be used to manage the data processing in computing resources [RFC2904].¶

  • Hardware-rooted attestation (e.g., TPM measurements) could be used for runtime authorization decisions.¶

• Resilience Against DoS¶

  • Proof-of-work challenges for request authentication could be deployed as the resilience against DoS during traffic anomalies.¶

  • Geo-distributed traffic scrubbing could be enabled through collaboration with CDN providers.¶

• Continuous Monitoring¶

  • Nodes could be instrumented with runtime integrity verification using OpenTelemetry standards.¶

  • Anomaly detection systems leveraging federated learning could be established to identify cross-node attack pattern.¶

4. Computing Path Selector Security Issues

The Computing Path Selector which is responsible for dynamically selecting optimal forwarding paths, faces the following threats:¶

• Path Manipulation Attacks¶

  Adversaries may forge or alter path metadata (e.g., node capabilities, network latency) to steer computation tasks toward compromised nodes.¶

• Covert Channel Exploitation¶

  Path selection patterns could be abused to leak sensitive information through timing analysis or topology fingerprinting.¶

• Topology Poisoning¶

  Injection of forged network topology data could degrade path selection efficiency or enable man-in-the-middle (MITM) attacks.¶

• Decision Logic Corruption¶

  Runtime modification of C-PS algorithms may lead to suboptimal or adversarial path selections.¶

• Orchestrator Impersonation¶

  Spoofed control-plane messages could trick CPS into accepting unauthorized path directives.¶

To mitigate these risks, CATS implementations COULD implement the following countermeasures:¶

• Authenticated Path Metadata¶

  • Digitally sign topology updates and node capability information could be implemented using CBOR Object Signing and Encryption (COSE) [RFC9052].¶

  • Enforce strict schema validation for path attributes per IETF YANG models [RFC7950].¶

• Decision Integrity Protection¶

  • C-PS path selection logic could be isolated in hardware-rooted trusted execution environments (TEEs).¶

  • Runtime attestation of decision engines could be implemented via Remote Attestation Procedures (RATS) [RFC9334].¶

• Differential Privacy for Path Selection¶

  • Sensitive selection patterns could be Obfuscated by incorporating differentially private noise.¶

• Resilient Topology Discovery¶

  • RPKI [RFC6480] or BGPsec principles [RFC8205] could be adopted for secure topology propagation in multi-domain scenarios.¶

• Control-Plane Hardening¶

  • Mutual authentication could be adopted in communications between C-PS and C-SMA or C-NMA via OAuth 2.1 [RFC9449].¶

5. Computing Service Announcement Security Issues

The announcement of computing services in distributed environments introduces several security risks that must be addressed to ensure system integrity, confidentiality, and availability. This section outlines key threats and proposed countermeasures.¶

• Unauthorized Announcement Injection¶

  Malicious actors may forge or manipulate service announcements to advertise rogue computing nodes, redirect traffic to compromised resources, or disrupt service discovery, which may lead to data exfiltration, computation tampering or denial of service.¶

• Sensitive Information Exposure¶

  Service announcements containing unencrypted metadata (e.g., topology details, capability descriptors) may reveal sensitive infrastructure or operational patterns, which may lead to attack surface expansion for targeted exploits or reconnaissance.¶

• Replay/Reuse of Legacy Announcements¶

  Replayed announcements of deprecated services could lead to resource misallocation or dependency on outdated/insecure compute nodes.¶

• DoS Through Announcement Flooding¶

  Flooding the control plane with excessive or malformed announcements may lead to system resources exhausted.¶

• Identity Spoofing¶

  Impersonation of legitimate service providers through forged identity claims in announcements.¶

To address these risks, CATS implementations COULD adopt the following mitigation measures:¶

• Cryptographic Integrity Protection¶

  • Digital signatures (e.g., using COSE/JOSE) could be adopted for all announcements to ensure authenticity and integrity.¶

  • Verifiable attestation (via frameworks like RATS) could be used for critical service claims.¶

• Metadata Minimization & Encryption¶

  • Data minimization principles could be applied to limit exposed metadata in announcements.¶

  • Hybrid encryption (e.g., ECIES) could be used for sensitive fields while maintaining routable/public attributes in cleartext.¶

• Anti-Replay Mechanisms¶

  • Timestamp/nonce could be used in announcements with strict freshness validation.¶

• Rate Limiting & Prioritization¶

  • QoS controls could be applied to prioritize announcements from authenticated entities.¶

  • Rate limits per node/domain could be adopted using token-bucket or similar algorithms.¶

• Identity Verification¶

  • The announcement from the computing devices could be binded to DIDs (Decentralized Identifiers) or VCs (Verifiable Credentials) for cryptographic identity proof.¶

6. Metrics Distribution Security Issues

Metrics distribution mechanisms in CATS are critical for performance optimization and resource coordination. However, they introduce specific security challenges that must be mitigated to prevent misuse or systemic compromise. This section identifies key threats and proposes countermeasures.¶

• Tampering with Metric Data¶

  Adversaries may alter metrics (e.g., latency, throughput, resource utilization) during transmission to mislead the decision-making of control plane, triggering suboptimal traffic placement or resource allocation and leading to degraded service performance.¶

• Eavesdropping on Sensitive Metrics¶

  Unauthorized interception of metrics may cause the eavesdropping on sensitive operational details (e.g., geo-location patterns, infrastructure capacity), which will lead to the exposure of business-critical intelligence or user behavior profiling.¶

• Forged Metric Sources¶

  Spoofing of metric publishers to inject false data or impersonate trusted entities (e.g., fake "low-load" signals to attract traffic).¶

• Privacy Violations via Aggregation¶

  The statistical analysis of aggregated metrics may produce inference of sensitive information (e.g., user activity, infrastructure weaknesses) which may result in privacy violation.¶

To address these risks, CATS implementations COULD adopt the following safeguards:¶

• End-to-End Integrity Protection¶

  • Cryptographic signatures (e.g., using COSE/JOSE) could be applied to metric payloads to ensure authenticity and detect tampering.¶

  • Hash-chaining or Merkle trees could be used for batch metric verification in streaming scenarios.¶

• Confidentiality Preservation¶

  • Sensitive metric fields could be encrypted (e.g., using AES-GCM or HPKE)while preserving routable headers in cleartext.¶

  • Differential privacy or noise injection could be employed for aggregated metrics to prevent inference attacks.¶

• Source Authentication¶

  • Metric publishers could be binded to cryptographically verifiable identities (e.g., X.509 certificates, DIDs).¶

  • Role-based access control (RBAC) could be used for metric publication rights.¶

• Privacy-Aware Metric Design¶

  • The high-granularity data (e.g., masking exact geolocation to regional levels) could be anonymized or truncated to protect the privacy.¶

  • The federated learning or on-device aggregation could be used to minimize raw data exposure.¶

7. Security-related Metrics

The service and network metrics could include the security-related capabilities which could be used by the CATS Path selector to compute paths with security guarantee.¶

The security capabilities of nodes could be one of the metrics for C-PS to computing the traffic forwarding path and form a secure routing path. And C-PS will fetch the real-time awareness of the security capabilities available in the network and computing services and finally provide security protection for users. Clients with high security requirements could choose the service with desired security attributes and achieve dependable forwarding on top of only devices that satisfies certain trust requirements, which will avoid the risks of traffic eavesdropping, sensitive data leakage etc.¶

8. Security Considerations

The security considerations of CATS are presented throughout this document. .¶

9. IANA Considerations

This document has no IANA actions.¶

10. References

Acknowledgements

TBD¶

Authors' Addresses