NASR Use Case and Requirements

1.1. Note for NASR participants

This document collates and synthesizes discussion outcomes of NASR mailing list and side meetings.¶

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 will stay as an informational draft.¶

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 authorities. NASR can help operator to attest to an orchestrated path and provide verifiable forwarding proofs to help clients/authorities audit the service.¶

SFC is used as an example, therefore network elements are not limited to Service Functions, and paths are not limited to a SFC path. Other devices or network functions may incorporate features (built-in security capabilities, roots of trust and attestation mechanisms, etc.) suitable to support path validation.¶

5.2. Use Case 2: Verifying Path Properties

In use case 1, the orchestrated path is explicit and specific down to each network element. Sometimes, clients do not need to know every detail of the network path. Rather, clients will request the verification of a certain property within the path, such as trustworthiness, security level, geolocation, vendor characteristics, transit provider, etc. from the operator. Using NASR, the operator can orchestrate this path by selecting network elements and links with the requested properties, attest to the path, and verifiably prove to clients the path properties and that the traffic did follow this path.¶

In both this and the previous case, the order of the elements in the path may not be important, as the requests may be limited to a set of attributes for the path nodes, or the guarantee that traffic traversed a certain (set of) node(s).¶

5.3. Use Case 3: Sensitive Data Routing

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

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: Sensitive data transmission due to remote AI training

This use case is similar to Use Case 3 but is more specific. As AI trend rise, operators are investing in "AI training centers" for lease. Due to scalability and cost reduction considerations, training centers tend to be built separately from data centers. In manufacturing industry or other data-heavy industries, DCs or private storage is often built next to the campus. But in order to support training and utilize operator-running training centers, wide-area data transmission between DC and TC is needed. Enterprise clients, when faced with privacy-sensitive data leaving their DCs and go through wide-area transmission, are highly unhappy. Yet, it is also impractical for operators to build dedicated training centers next to the client DC. Without NASR guaranteeing dependable forwarding and non-leakage, the market sales of operator's training center business is hindered.¶

5.5. Use Case 5: Ingress Filtering

Ingress Filtering techniques help prevent source IP address spoofing and denial-of-service (DoS) attacks [RFC8704][RFC5635]. Approches like uRPF 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. Furthermore, when uRPF is not available and source address cannot be trusted, NASR can offer a way to filter malicious traffic based on the path used to carry out such an attack [Yaar03]. The other usage is to check if a packet carries a valid trail of transit proofs. If it does then the packet is verified.¶

6.1. Requirement 1: Attributes of a network element, interfaces

According to goal 1 of NASR definition, NASR (to-be) Working Group will define security/trustworthiness attributes of network elements, for clients to request from operators. Attributes should be objective claims, including but not limited to existing remotely-attested claims, element type (physical or virtual network element), security capability it enables, cryptographic algorithms, key strength, deployed geolocation, 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 (Hardware Model)¶

• hwversion (Hardware Version)¶

• swname (Software Name )¶

• swversion (Software Version)¶

• location (location)¶

Some new claim extensions can be made:¶

elemtype
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.¶

Additionally, service request interface between clients and operators should also be defined.¶

6.2. Requirement 2: Path Attestation Procedures

After a path attribute request is sent from the client to the operator, the operator will have to choose qualifying devices and orchestrate a path. According to the goal 2 of NASR definition, the path attestation procedure will be the core protocol of NASR. The procedure should be able to re-attest to the whole path and provide path-level attribute proofs (attestation results), proving the delivered path complies with client request, which should reuse RATS procedures. Details are being designed in the NASR architecture document.¶

6.3. Requirement 3: 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 transited certain network elements, and it can include a verification of the order in which those elements where transited (Ordered POT, OPoT) or not. A secure POT mechanism should verifiably reflect the identity of the transited network elements and their relevant attributes, if applicable:¶

• For basic POT, there is no further attribute than the identity of the transited element and, optionally, its relative position/order within the path. This is the goal of the POT mechanism defined in [I-D.ietf-sfc-proof-of-transit-08].¶

• For extended POT, different attributes can be considered from a list of relevant ones: trustworthiness measure, available security capabilities, geolocation, vendor, etc. This needs the definition of the relevant attributes of a network element, which is discussed in Section 6.1¶

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 aggregated into one collective summary at 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.¶

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, essentially, the opposite to Req. 1 Proof-of-Transit. Certain potential user have expressed their interest on 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, especially when Ordered POT (OPOT) is enforced. 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 remotely attested 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 the 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 further stages of NASR, in case of a rechartering.¶

8.1. Initially targeting for intra-domain or inter-domain scenario?

Limited domain with some trust assumptions and controls to devices will be easy to start with, but inter-domain scenario will be critical for standardization.¶

8.2. Does tunneling solve the problem?

Tunnels, VPNs do not perceive the underlying network devices. Quality measurements can be done, but other detail information of bearing devices are not visible.¶

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

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