1.1. Requirements Notation

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶

2.1. Algorithms

The algorithm field is not discussed by this document. Readers are encouraged to read [RFC8624] for guidance about DNSSEC algorithm usage.¶

2.2. Flags

The NSEC3PARAM flags field currently contains only reserved and unassigned flags. However, individual NSEC3 records contain the "Opt-Out" flag [RFC5155] that specifies whether that NSEC3 record provides proof of nonexistence. In general, NSEC3 with the Opt-Out flag enabled should only be used in large, highly dynamic zones with a small percentage of signed delegations. Operationally, this allows for fewer signature creations when new delegations are inserted into a zone. This is typically only necessary for extremely large registration points providing zone updates faster than real-time signing allows or when using memory-constrained hardware. Operators considering the use of NSEC3 are advised to carefully weigh the costs and benefits of choosing NSEC3 over NSEC. Smaller zones, or large but relatively static zones, are encouraged to not use the opt-opt flag and to take advantage of DNSSEC's authenticated denial of existence.¶

2.3. Iterations

NSEC3 records are created by first hashing the input domain and then repeating that hashing using the same algorithm a number of times based on the iteration parameter in the NSEC3PARAM and NSEC3 records. The first hash with NSEC3 is typically sufficient to discourage zone enumeration performed by "zone walking" an unhashed NSEC chain.¶

Note that [RFC5155] describes the Iterations field as follows¶

The Iterations field defines the number of additional times the hash function has been performed.¶

This means that an NSEC3 record with an Iterations field of 0 actually requires one hash iteration.¶

Only determined parties with significant resources are likely to try and uncover hashed values, regardless of the number of additional iterations performed. If an adversary really wants to expend significant CPU resources to mount an offline dictionary attack on a zone's NSEC3 chain, they'll likely be able to find most of the "guessable" names despite any level of additional hashing iterations.¶

Most names published in the DNS are rarely secret or unpredictable. They are published to be memorable, used and consumed by humans. They are often recorded in many other network logs such as email logs, certificate transparency logs, web page links, intrusion-detection systems, malware scanners, email archives, etc. Many times a simple dictionary of commonly used domain names prefixes (www, mail, imap, login, database, etc.) can be used to quickly reveal a large number of labels within a zone. Because of this, there are increasing performance costs yet diminishing returns associated with applying additional hash iterations beyond the first.¶

Although Section 10.3 of [RFC5155] specifies the upper bounds for the number of hash iterations to use, there is no published guidance for zone owners about good values to select. Recent academic studies have shown that NSEC3 hashing provides only moderate protection [GPUNSEC3] [ZONEENUM].¶

2.4. Salt

NSEC3 records provide an additional salt value, which can be combined with a Fully Qualified Domain Name (FQDN) to influence the resulting hash, but properties of this extra salt are complicated.¶

In cryptography, salts generally add a layer of protection against offline, stored dictionary attacks by combining the value to be hashed with a unique "salt" value. This prevents adversaries from building up and remembering a single dictionary of values that can translate a hash output back to the value that it was derived from.¶

In the case of DNS, the situation is different because the hashed names placed in NSEC3 records are always implicitly "salted" by hashing the FQDN from each zone. Thus, no single pre-computed table works to speed up dictionary attacks against multiple target zones. An attacker is always required to compute a complete dictionary per zone, which is expensive in both storage and CPU time.¶

To understand the role of the additional NSEC3 salt field, we have to consider how a typical zone walking attack works. Typically, the attack has two phases: online and offline. In the online phase, an attacker "walks the zone" by enumerating (almost) all hashes listed in NSEC3 records and storing them for the offline phase. Then, in the offline cracking phase, the attacker attempts to crack the underlying hash. In this phase, the additional salt value raises the cost of the attack only if the salt value changes during the online phase of the attack. In other words, an additional, constant salt value does not change the cost of the attack.¶

Changing a zone's salt value requires the construction of a complete new NSEC3 chain. This is true both when re-signing the entire zone at once and when incrementally signing it in the background where the new salt is only activated once every name in the chain has been completed. As a result, re-salting is a very complex operation, with significant CPU time, memory, and bandwidth consumption. This makes very frequent re-salting impractical and renders the additional salt field functionally useless.¶

3.1. Best Practice for Zone Publishers

First, if the operational or security features of NSEC3 are not needed, then NSEC SHOULD be used in preference to NSEC3. NSEC3 requires greater computational power (see Appendix B) for both authoritative servers and validating clients. Specifically, there is a nontrivial complexity in finding matching NSEC3 records to randomly generated prefixes within a DNS zone. NSEC mitigates this concern. If NSEC3 must be used, then an iterations count of 0 MUST be used to alleviate computational burdens. Note that extra iteration counts other than 0 increase the impact of CPU-exhausting DoS attacks, and also increase the risk of interoperability problems.¶

Note that deploying NSEC with minimally covering NSEC records [RFC4470] also incurs a cost, and zone owners should measure the computational difference in deploying either [RFC4470] or NSEC3.¶

In short, for all zones, the recommended NSEC3 parameters are as shown below:¶

; SHA-1, no extra iterations, empty salt:
;
bcp.example. IN NSEC3PARAM 1 0 0 -

For small zones, the use of opt-out-based NSEC3 records is NOT RECOMMENDED.¶

For very large and sparsely signed zones, where the majority of the records are insecure delegations, opt-out MAY be used.¶

Operators SHOULD NOT use a salt by indicating a zero-length salt value instead (represented as a "-" in the presentation format).¶

If salts are used, note that since the NSEC3PARAM RR is not used by validating resolvers (see Section 4 of [RFC5155]), the iterations and salt parameters can be changed without the need to wait for RRsets to expire from caches. A complete new NSEC3 chain needs to be constructed and the full zone needs to be re-signed.¶

3.2. Recommendation for Validating Resolvers

Because there has been a large growth of open (public) DNSSEC validating resolvers that are subject to compute resource constraints when handling requests from anonymous clients, this document recommends that validating resolvers reduce their iteration count limits over time. Specifically, validating resolver operators and validating resolver software implementers are encouraged to continue evaluating NSEC3 iteration count deployment trends and lower their acceptable iteration limits over time. Because treating a high iterations count as insecure leaves zones subject to attack, validating resolver operators and validating resolver software implementers are further encouraged to lower their default limit for returning SERVFAIL when processing NSEC3 parameters containing large iteration count values. See Appendix A for measurements taken near the time of publication of this document and potential starting points.¶

Validating resolvers MAY return an insecure response to their clients when processing NSEC3 records with iterations larger than 0. Note also that a validating resolver returning an insecure response MUST still validate the signature over the NSEC3 record to ensure the iteration count was not altered since record publication (see Section 10.3 of [RFC5155]).¶

Validating resolvers MAY also return a SERVFAIL response when processing NSEC3 records with iterations larger than 0. Validating resolvers MAY choose to ignore authoritative server responses with iteration counts greater than 0, which will likely result in returning a SERVFAIL to the client when no acceptable responses are received from authoritative servers.¶

Validating resolvers returning an insecure or SERVFAIL answer to their client after receiving and validating an unsupported NSEC3 parameter from the authoritative server(s) SHOULD return an Extended DNS Error (EDE) [RFC8914] EDNS0 option of value 27. Validating resolvers that choose to ignore a response with an unsupported iteration count (and that do not validate the signature) MUST NOT return this EDE option.¶

Note that this specification updates [RFC5155] by significantly decreasing the requirements originally specified in Section 10.3 of [RFC5155]. See the Security Considerations (Section 4) for arguments on how to handle responses with non-zero iteration count.¶

3.3. Recommendation for Primary and Secondary Relationships

Primary and secondary authoritative servers for a zone that are not being run by the same operational staff and/or using the same software and configuration must take into account the potential differences in NSEC3 iteration support.¶

Operators of secondary services should advertise the parameter limits that their servers support. Correspondingly, operators of primary servers need to ensure that their secondaries support the NSEC3 parameters they expect to use in their zones. To ensure reliability, after primaries change their iteration counts, they should query their secondaries with known nonexistent labels to verify the secondary servers are responding as expected.¶