NASR Use Case and Requirements

1.1. Note for NASR participants

This document collates and synthesizes discussion outcomes of NASR mailing list and IETF 118 path validation side meeting.¶

It is created to help 1. Foster consensus among list members. 2. Orient non-list members to NASR goals and current progress¶

This document may become a WG-draft but stay informational.¶

5.1. Use Case 1: Network Path Validation

Explicit routing protocols permit explicit control over the traffic path, in order to meet certain performance, security or compliance requirements. For example, operators can use SRv6 to orchestrate a Service Function Chaining (SFC) path and provide packaged security services or compliance services. For either of them, validating the actual traffic path in the forwarding plane as an auditing measure is needed for clients and/or authorties. NASR can help operator to attest to an orchestrated path and provide verifiable forwarding proofs to help clients/authorities audit the service.¶

The network element is not limited to Service Function-- it can also be devices that has certain built-in security capabilities (or other attributes), or workloads. Hence the path is not limited to a SFC path.¶

5.2. Use Case 2: Verifying Path Properties

In use case 1, The orchestrated path is explict and specfic down to each network element. Sometimes, the client does not need to know every detail. Rather, the clients will request a path of a certain property, such as trustworthiness, security level, location, vendor, etc, from the operator. With NASR, the operator can orchestrate this path by selecting network elements with requested properties, attest to the path, and verifiably prove to the clients the traffic did follow this path.¶

Compared to the first use case, the order of the elements may not be important. This use case is more focused on validating the attributes of the path.¶

5.3. Use Case 3: Sensitive Data Routing

Clients from specific industries such as finance, governments have very low tolerance of data leakage. These clients require assurance that their data only travels on top of their selected leased line, MPLS VPN or SD-WAN path, and have (preferably real-time) visibility evidence or proof. Some compliance requirements also prohibit customer data escape a specific geolocation without permission. To avoid data leakage risks and compliance risks, some clients are willing to pay a premium for high data routing security guarantees. NASR can detect for such violations and make corrections promptly.¶

Compared to the first and second use case, this use case also requires some preventive measures before a wrongful forwarding happens at the first place.¶

5.4. Use Case 4: Ingress Filtering

Ingress Filtering techniques, such as uRPF, help prevent source IP address spoofing and denial-of-service (DoS) attacks [RFC8704][RFC5635]. It works by validating the source IP address of a received packet by performing a reverse path lookup in FIB table, all the way to the source. If the path does not exist, the packet is dropped. NASR can be used to regularly validate the path stored in the FIB table, and tell if it continues to exist. This can potentially reduce the false negative rate.¶

6.1. Requirement 1: Proof-of-Transit (POT) Mechanisms

All use cases requested public verifiability of packet transit history. Proof-of-Transit (POT) is a proof that a packet DID transit certain network elements. A secure POT mechanism should truthfully reflect the identity of the network element and its attributes. The "attribute" could be different:¶

• For simple POT, the "attribute" means the path it is on, and its relative position/order on the path. This is the goal of POT mechanism defined in [I-D.ietf-sfc-proof-of-transit-08].¶

• For richer POT, the "attribute" means it could be a list of attributes: trustworthiness, security capabilities it has, geolocation, vendor, etc. This needs the definition of attributes of a network element, which is discussed in Section 6.2¶

According to use case 2, the granularity of POT may also differ. POT can be generated and recorded on a per-hop basis, or can be merged into one collective summary in the path level.¶

The most appropriate POT mechanism for each scenarios may differ-- inter-domain or intra-domain, with or without a pre-attest, per-packet or on-demand, privacy-preserving or not, etc.¶

6.2. Requirement 2: Attributes of a network element

The identity of a subject should be defined by the attributes (or claims) it owns. Attribute-defined identity is a paradigm widely accepted in SCIM [RFC7643], OAuth [RFC7519], SAML [SAML2], etc. POT proof should reflect the identity and associated attributes, such as element type, security level, security capability it has, remote-attestated or not, vendor, deployed geolocation, current timestamp, path it is on, hop index on the path etc.¶

Such attributes/claims/attestation results can reuse existing specifications, for example [I-D.ietf-rats-eat], [I-D.ietf-rats-ar4si] in RATS WG. Some existing claims that we can reuse:¶

hwmodel
hwversion
swname
swversion
location

Some new claim extensions can be made:¶

elemtype
pathid
index
secfunctions
vendor
...

(subject to discussion, add, change)¶

NASR could work closely with RATS on the standardization of above attributes and means of proving them.¶

6.3. Requirement 3: Path Attestation Procedures

After a path is selected, it should be¶

1. commited to prevent changes,¶

2. publicized for common referencing and retrival.¶

The stored path should contain these information: univeral ID, all network elements on the path, and attributes of them. (Schemas may vary depending on scenarios)¶

TBA¶

7.1. Non-Requirements 1: Proof-of-Non-Transit (PONT) Mechanisms

Proof-of-Non-Transit (PONT) is a proof that a packet did NOT transit certain network elements. It is to the opposite to the Req. 1 Proof-of-Transit. Certain customers requested PONT for compliance or security purposes.¶

First of all, PONT is a non-inclusion proof, and such non-existence proof cannot be directly given. Second, under certain circumstances, PONT can be inferred from POT. For example, assume devices are perfectly secure and their behaviors completely compliant to expectations, then POT over A-B-C indicates the packet did not transit X/Y/Z. To relax the security assumptions, if the devices are remote attestated and such claim is proved by POT, then the packet should only transited these trusted devices, assuming the RATS procedure is secure. The reliability of such reasoning decreases as the security level of device decreases.¶

NASR mailing list has agreed NOT to provide PONT mechanisms, but could provide some informational measures and conditions that could indicate PONT from POT results. For example, under xxx constraints and circumstances, if traffic passed X AND Y (device or geolocation), then it did NOT (or with a quantifiably low probability it did not) pass Z.¶

Since this part is research-related, NASR will work with PANRG and Academia for counseling.¶

7.2. Future Requirement 2: Packet Steering and Preventive Mechanisms

In sensitive data routing use case, it is certainly necessary to know and verify the transit path of a packet using POT mechanisms. However, it might be too late to have the data already exposed to the insecure devices and risk leakage. There should be packet steering mechanisms or other preventive measures that help traffic stay in the desired path. For example, doing an egress check before sending to the next hop, preventing sending packet to a device with a non-desirable attribute.¶

The mailing list and side meeting has received requests to this requirement, it should fall in NASR interest, but also agreed this may not be part of the initial scope of NASR-- it is a topic to be included in the next stage of NASR when rechartering.¶

8.1. Why not use static routing?

Static routing severely limits the scalability and flexibility for performance optimizations and reconfigurations. Flexible orchestration of paths will be prohibited. Also, even static routing is used, we still need proof of transit for compliance check.¶

8.2. Initially targeting for intradomain or interdomain scenario?

8.3. Does tunneling solve the problem?

8.4. Does all nodes on the path need to compute the POT?

Whether the validation is strict or loose is dependent on the scenario. For example in SFC use case, we are only interested in verifying some important elements of interes processed the traffic.¶