SRv6 Security¶
Security is a critical aspect of any SRv6 deployment. This page covers the security mechanisms built into SRv6, common attack vectors, and best practices for securing an SRv6 network.
Threat Model¶
What Could Go Wrong?¶
| Threat | Description | Impact |
|---|---|---|
| SRH injection | Attacker injects packets with forged SRH to steer traffic through unintended paths | Traffic interception, bypass of security functions |
| SRH manipulation | Attacker modifies the segment list in transit to alter the forwarding path | Traffic diversion, service chain bypass |
| SID probing | Attacker probes the SRv6 SID space to discover network topology | Reconnaissance for further attacks |
| Amplification | Attacker uses SRv6 encapsulation to amplify traffic volume | DDoS amplification |
| Bypass service chain | Attacker crafts packets that skip security functions in a service chain | Evading firewall/IDS inspection |
Defense Mechanisms¶
1. Infrastructure ACLs (iACLs)¶
The first and most important defense: filter SRv6 packets at the network boundary.
SRv6 Filtering Rules¶
┌──────────────────┐
External ──────►│ Border Router │──────► Internal SRv6 Domain
Traffic │ │
│ iACL: │
│ DENY SRH from │
│ outside │
└──────────────────┘
Key principle: SRv6 SRH should never be accepted from outside your network. Only traffic originating from within your SRv6 domain should carry an SRH.
2. SRH HMAC (RFC 8754)¶
The SRH supports an HMAC TLV that provides cryptographic authentication of the segment list:
How HMAC Works¶
- Ingress node computes HMAC over the segment list using a pre-shared key
- HMAC value is inserted as a TLV in the SRH
- Each SRv6 node (optionally) verifies the HMAC before processing the SID
- If HMAC validation fails, the packet is dropped
What HMAC Protects Against¶
| Protection | Description |
|---|---|
| SRH forgery | Attacker cannot create valid HMAC without the key |
| SRH tampering | Any modification to the segment list invalidates the HMAC |
HMAC does not provide replay protection. RFC 8754 defines no sequence number or timestamp in the HMAC TLV, so a captured packet with a valid HMAC is still valid if it is resent.
HMAC TLV Format¶
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Type (5) | Length |D| Reserved |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| HMAC Key ID (4 bytes) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
| HMAC Value (32 bytes) |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
HMAC tradeoffs
HMAC adds 40 bytes of overhead per packet (with a 32-byte HMAC) and requires key management across all SRv6 nodes. Most deployments rely on iACLs at the boundary rather than per-packet HMAC for performance reasons.
3. SID Block Filtering¶
Restrict the SRv6 SID block to your network. For new allocations, use the IANA SRv6 SID block 5f00::/16 (RFC 9602): it is registered as not globally reachable and is easy to recognize in filters. The examples on this page use 5f00::/16; if your locators come from your own GUA or another range (for example fcbb:bb00::/24, common in lab configurations), substitute that prefix.
Keep SIDs inside the SR domain at two layers (see SID Structure):
- Routes: never advertise the SID block to external BGP peers, and never accept it from them
- Packets: drop packets arriving on external interfaces whose destination is in the SID block
- uRPF: Enable Unicast Reverse Path Forwarding to catch spoofed SRv6 source addresses
4. Rate Limiting¶
Rate-limit SRv6-related control plane traffic:
| What to Rate Limit | Why |
|---|---|
| ICMPv6 for SRv6 SIDs | Prevent ping/traceroute-based recon floods |
| SRH packets to control plane | Prevent CPU exhaustion from SRH processing |
| BGP SRv6 SID updates | Prevent BGP flooding attacks |
Security Best Practices¶
Deployment Checklist¶
- iACLs on all external interfaces — drop packets with SRH (routing type 4) from outside your domain
- SID block not advertised externally — BGP policy to filter SRv6 locator prefixes
- uRPF enabled — prevents spoofed source addresses in SRv6 packets
- Rate limiting on SRv6-related ICMP and control plane traffic
- My SID table validation — nodes only process SIDs they are configured to own
- Monitoring — alert on unexpected SRH packets at boundary routers
- Key rotation plan if using HMAC (for high-security environments)
Domain Trust Model¶
SRv6 operates on a trusted domain model:
┌─────────────────────────────────────┐
│ SRv6 Trusted Domain │
│ │
│ All nodes trust each other's SIDs │
│ SRH is valid within this domain │
│ │
│ ┌────┐ ┌────┐ ┌────┐ │
│ │ PE1│───│ P1 │───│ PE2│ │
│ └────┘ └────┘ └────┘ │
│ │
├──────────────────────────────────────┤
│ Border: iACLs drop external SRH │
└──────────────────────────────────────┘
│ │
External traffic External traffic
(no SRH allowed) (no SRH allowed)
This is the same trust model used by MPLS networks today — MPLS labels are not validated, only trusted within the operator's domain. SRv6 improves on this by offering optional HMAC for environments requiring stronger guarantees.
SRH Security Analysis (RFC 8754 Sections 5 and 7)¶
RFC 8754 Section 5 (the intra-SR-domain deployment model) and Section 7 (security considerations), together with RFC 8402 Section 8, set out the security model for SRv6. Understanding the threat categories and deployment models is essential for designing a secure SRv6 network.
Threat Categories¶
Three threat categories follow from those sections:
| Threat Category | Description | RFC Reference |
|---|---|---|
| Source address spoofing | An attacker outside the trusted domain crafts packets with a spoofed source address belonging to the SRv6 domain, causing them to bypass iACL filters that match on source prefix | RFC 8754 Sections 5.1 and 7 |
| SRH manipulation | An attacker intercepts packets in transit and modifies the segment list, Segments Left field, or other SRH fields to alter the forwarding path | RFC 8754 Sections 2.1.2 and 7 (HMAC) |
| Topology disclosure | An attacker probes the SRv6 SID space (which uses routable IPv6 addresses) to discover internal network topology, node roles, and service chain structure | RFC 8402 Section 8.2 |
Key insight from the RFCs
Unlike MPLS labels (which are locally significant and not routable), SRv6 SIDs are routable IPv6 addresses inside the domain. This makes topology discovery easier for attackers unless SID prefixes are properly filtered at domain boundaries.
Three Ways to Deploy SRv6 Security¶
RFC 8754 defines one deployment model, a trusted SR domain protected at its boundary (Section 5.1), and the HMAC TLV for cases where that is not enough (Section 2.1.2). In practice that gives three options, with increasing protection:
Model 1: Trusted Domain with Filtering¶
The most common deployment model. All nodes within the SRv6 domain are trusted. Security relies entirely on filtering at domain boundaries.
Untrusted ┌─────────────────────────────┐ Untrusted
Internet ────────►│ iACL: drop SRH (RT type 4) │◄──────── Internet
│ uRPF: reject spoofed src │
│ │
│ Trusted SRv6 Domain │
│ (no per-packet auth) │
└─────────────────────────────┘
- When to use: Single-operator networks where all routers and links are physically secured. This covers the vast majority of SP and enterprise deployments.
- Tradeoffs: Zero per-packet overhead, no key management, but offers no protection against compromised nodes inside the domain or man-in-the-middle attacks on internal links.
Model 2: HMAC-Based Verification¶
Every SRH carries a cryptographic HMAC TLV. Nodes verify the HMAC before processing SIDs, providing per-packet authentication of the segment list.
- When to use: Multi-domain environments, inter-provider SRv6 peering, or networks where internal links traverse untrusted infrastructure (e.g., leased dark fiber, shared data center interconnects).
- Tradeoffs: Strong per-packet security, but adds 40 bytes of overhead per packet, requires HMAC computation at every verifying node, and demands a key management infrastructure across all participating nodes.
Model 3: Combined (Filtering + HMAC)¶
Uses iACL filtering at domain boundaries and HMAC verification for traffic crossing trust boundaries (e.g., between autonomous systems or between the core and a less-trusted access layer).
- When to use: Large-scale networks with mixed trust zones, where some segments of the infrastructure are more exposed than others.
- Tradeoffs: Best overall security posture, but highest operational complexity. Requires careful definition of trust boundaries and selective HMAC application.
Practical guidance
Almost all production SRv6 deployments today use Model 1 (trusted domain with filtering). HMAC is specified in the standards but is rarely deployed due to performance and operational overhead. The industry consensus is that robust boundary filtering provides sufficient security for single-operator networks.
HMAC Deep Dive¶
The HMAC TLV defined in RFC 8754 Section 2.1.2 provides optional per-packet cryptographic authentication of the SRH. This section covers the computation details, key management challenges, and real-world deployment status.
HMAC Computation Details¶
Per RFC 8754 Section 2.1.2.1, the HMAC is computed over a concatenation of the following fields, as they would be received at the verifying node:
| Field | Size | Purpose |
|---|---|---|
| Source IPv6 Address | 16 bytes | Binds the HMAC to the packet originator, preventing source spoofing |
| Last Entry | 1 byte | Index of the last element of the segment list |
| Flags | 1 byte | SRH flags field |
| HMAC 16 bits following Length | 2 bytes | The D bit and Reserved field of the HMAC TLV |
| HMAC Key ID | 4 bytes | Identifies which pre-shared key and algorithm were used |
| Full Segment List | 16 x n bytes | All addresses in the segment list |
The hash algorithm is identified by the Key ID; implementations that support HMAC must support SHA-256. The digest is truncated to 32 bytes. With Type (1 byte), Length (1), D bit + Reserved (2) and Key ID (4), the total HMAC TLV overhead is 40 bytes.
HMAC = HMAC(Key, ! algorithm identified by the Key ID
IPv6-Source-Address ||
Last-Entry ||
Flags ||
D-bit+Reserved ||
HMAC-Key-ID ||
Segment-List[0..n]
)
What the HMAC does NOT cover
The HMAC does not protect the payload, the Segments Left field (which changes at each segment endpoint), the IPv6 destination address, or any SRH TLVs other than the fields listed above. Instead of hashing the destination address, the verifying node checks that it equals the current segment (Segment List[Segments Left]). It authenticates only the segment list and its associated metadata.
Key Management Challenges¶
HMAC-based SRH verification introduces significant key management requirements:
| Challenge | Description |
|---|---|
| Key distribution | Every node that needs to verify the HMAC must possess the correct pre-shared key. In a network with hundreds of routers, this requires a scalable key distribution mechanism. |
| Key rotation | Keys must be rotated periodically. During rotation, nodes must support at least two active Key IDs simultaneously (old and new) to avoid packet drops during the transition window. |
| Per-domain vs. per-node keys | A single domain-wide key is simpler to manage but means a compromise of any one node exposes the key for the entire domain. Per-node keys provide better isolation but multiply the management burden by the number of nodes. |
| No standard protocol | RFC 8754 does not specify a key distribution protocol. Operators must rely on out-of-band mechanisms (e.g., configuration management systems, proprietary key servers). |
Performance Impact¶
| Impact Area | Details |
|---|---|
| Packet overhead | 40 bytes per packet (HMAC TLV). For a 10-SID segment list, the SRH is already 168 bytes; HMAC adds another 24% overhead. |
| Computation cost | HMAC-SHA-256 computation at ingress (generation) and at each verifying node. On modern ASIC-based routers, this may require packet recirculation through the forwarding pipeline, effectively halving throughput for HMAC-verified traffic. |
| Latency | HMAC verification adds per-hop latency. For latency-sensitive traffic (e.g., 5G URLLC), this may be unacceptable. |
| ASIC support | Not all forwarding ASICs support HMAC computation in hardware. Software-based HMAC verification limits throughput to control-plane speeds. |
HMAC in Practice¶
Real-world deployment status
Despite being specified since RFC 8754 (March 2020), HMAC is rarely deployed in production. The vast majority of SRv6 operators rely exclusively on infrastructure ACLs (iACLs) at domain boundaries. Reasons include:
- Performance overhead is unacceptable at 100G/400G line rates without dedicated ASIC support
- Key management complexity across large router fleets
- The trusted domain model (with proper filtering) is considered sufficient for single-operator networks
- No interoperable multi-vendor HMAC implementation has been widely validated
HMAC remains valuable as a standards-based option for high-security environments or future multi-domain SRv6 interconnection scenarios.
SRv6-Specific Attack Vectors¶
The IETF draft draft-ietf-spring-srv6-security catalogs attack vectors specific to SRv6 deployments. Understanding each attack, its mitigation, and residual risk is critical for security planning.
1. SRH Insertion Attack¶
| Aspect | Details |
|---|---|
| Attack | An external attacker crafts an IPv6 packet with an SRH containing valid-looking SIDs and sends it into the SRv6 domain. If the packet is accepted, the attacker can steer traffic through arbitrary paths within the network. |
| Mitigation | Deploy iACLs on all external-facing interfaces to drop any packet containing an SRH (IPv6 Routing Header type 4). Combine with uRPF to reject packets with spoofed source addresses. |
| Residual risk | If any external-facing interface lacks the iACL filter, it becomes an entry point. Misconfigured or newly provisioned interfaces are the primary risk. Automation and compliance auditing of ACL deployment is essential. |
2. SID Spoofing¶
| Aspect | Details |
|---|---|
| Attack | An attacker (internal or external) forges packets with destination addresses matching valid SIDs (e.g., fcbb:bb00:0001:fe00:: for an End.DT4 function). If the SID is a valid entry in the target node's My SID table, the node will execute the associated behavior. |
| Mitigation | Edge filtering prevents external SID spoofing. Internally, use uRPF to validate that SRv6 source addresses originate from expected prefixes. The My SID table inherently limits processing to only configured SIDs. HMAC provides the strongest defense against SID spoofing. |
| Residual risk | A compromised internal node can generate packets with any valid SID. Without HMAC, there is no per-packet mechanism to detect internally originated spoofed SIDs. |
3. Service Chain Bypass¶
| Aspect | Details |
|---|---|
| Attack | An attacker crafts a segment list that deliberately omits security function SIDs (e.g., firewall, IDS/IPS) from the path. If a service chain is PE1 -> FW -> IDS -> PE2, the attacker constructs a segment list PE1 -> PE2, bypassing the security functions entirely. |
| Mitigation | Enforce service chaining at the headend via policy (e.g., SR Policy with explicit segment lists). Use HMAC to prevent modification of segment lists. Deploy monitoring to detect traffic arriving at egress PE nodes without having traversed mandatory service functions. |
| Residual risk | If the headend node itself is compromised, or if the SR Policy configuration is tampered with, service chains can be bypassed. Defense-in-depth (redundant enforcement points) reduces this risk. |
4. Amplification via Encapsulation¶
| Aspect | Details |
|---|---|
| Attack | An attacker sends a small packet to an SRv6 ingress node that encapsulates it with a long SRH (e.g., 10+ SIDs). The encapsulated packet is significantly larger (160+ bytes of SRH overhead), creating an amplification effect. If reflected or looped, this amplifies DDoS traffic volume. |
| Mitigation | Rate-limit SRv6 encapsulation for traffic from untrusted sources. Limit the maximum number of SIDs allowed in an SRH (policy enforcement at headend). Deploy iACLs to prevent external traffic from triggering SRv6 encapsulation. |
| Residual risk | Internal amplification remains possible if an attacker gains access to an internal node. The amplification factor is bounded by the maximum SRH size (theoretically up to ~137 SIDs in a standard IPv6 packet). |
5. Topology Discovery via Traceroute¶
| Aspect | Details |
|---|---|
| Attack | An attacker sends traceroute probes (ICMPv6 hop-limit exceeded) targeting addresses within the SRv6 SID block. By systematically probing the SID address space, the attacker can discover: active SIDs, node locators, function assignments, and service chain topology. |
| Mitigation | Rate-limit ICMPv6 responses for SRv6 SID prefixes. Filter ICMPv6 error messages at domain boundaries to prevent internal topology leakage. Do not advertise SRv6 locator prefixes externally. Consider using OAM-specific SIDs that are rate-limited separately. |
| Residual risk | Even with rate limiting, a patient attacker can slowly enumerate the SID space over time. Because SRv6 SIDs are IPv6 addresses, standard IPv6 scanning tools work against them. The 128-bit address space provides some protection, but locator prefixes typically use short prefix lengths (e.g., /48), narrowing the search space. |
Critical takeaway
The common thread across all these attacks is that boundary filtering is the primary defense. Every attack vector is significantly mitigated by properly deployed iACLs that prevent external SRH injection and SID block reachability.
Filtering Best Practices¶
This section expands on the iACL and filtering guidance with specific rules for different network boundaries.
Edge Filtering (External Interfaces)¶
All interfaces facing external networks (peering, customers, internet transit) must implement the following minimum filters:
!! Mandatory: Drop ALL packets with SRH from external sources
ipv6 access-list SRV6-EDGE-INGRESS
10 deny ipv6 any any routing-type 4
20 deny ipv6 any 5f00::/16 !! Drop traffic destined to SID block
30 deny ipv6 5f00::/16 any !! Drop traffic sourced from SID block
40 permit ipv6 any any
!
interface TenGigE0/0/0/0
description EXTERNAL-PEER
ipv6 access-group SRV6-EDGE-INGRESS ingress
!
!! Prevent SRv6 internal details from leaking outward
ipv6 access-list SRV6-EDGE-EGRESS
10 deny ipv6 any any routing-type 4 !! No SRH should leave the domain
20 deny ipv6 5f00::/16 any !! No SID-sourced packets leave
30 permit ipv6 any any
!
interface TenGigE0/0/0/0
description EXTERNAL-PEER
ipv6 access-group SRV6-EDGE-EGRESS egress
!
Inter-Domain Filtering¶
When SRv6 domains interconnect (e.g., between autonomous systems), additional filtering is required:
| Filter Rule | Direction | Purpose |
|---|---|---|
| Drop SRH with SIDs from the remote domain's SID block | Ingress | Prevent the remote domain from steering traffic through your internal paths |
| Drop SRH with SIDs from your own SID block | Egress | Prevent your internal SRH from leaking to the remote domain |
| Allow only agreed-upon inter-domain SIDs | Both | If inter-domain SRv6 is used, permit only explicitly negotiated SIDs |
SID Block Filtering via BGP¶
The SRv6 SID block (5f00::/16, or your own locator range) must never be advertised to external BGP peers, and should not be accepted from them either:
!! BGP policy to suppress SRv6 locator prefixes in both directions
prefix-set SRV6-SIDS
5f00::/16 le 128
end-set
!
route-policy DENY-SRV6-LOCATORS
if destination in SRV6-SIDS then
drop
endif
pass
end-policy
!
router bgp 65000
neighbor 198.51.100.1
address-family ipv6 unicast
route-policy DENY-SRV6-LOCATORS in
route-policy DENY-SRV6-LOCATORS out
!
!
!
Common misconfiguration
If SRv6 locator prefixes leak into the global routing table, any host on the internet can send packets directly to your SIDs. This bypasses the assumption that SID addresses are unreachable from outside the trusted domain and is the single most dangerous SRv6 misconfiguration.
uRPF Considerations for SRv6¶
Unicast Reverse Path Forwarding (uRPF) validates that the source address of incoming packets is reachable via the interface the packet arrived on.
| uRPF Mode | Behavior with SRv6 | Recommendation |
|---|---|---|
| Strict mode | Validates source address against the FIB for the specific ingress interface. May drop legitimate SRv6 traffic in asymmetric routing scenarios. | Use on access/edge interfaces where routing is symmetric. |
| Loose mode | Validates that the source address exists in the FIB (any interface). Allows asymmetric routing but provides weaker spoofing protection. | Use on core interfaces or where asymmetric routing is expected. |
| Feasible-path mode | Validates against all feasible paths (BGP best path + alternatives). Best balance for SRv6 core networks. | Preferred for SRv6 core interfaces where available. |
!! Enable strict uRPF on external interfaces
interface TenGigE0/0/0/0
description EXTERNAL-PEER
ipv6 verify unicast source reachable-via rx
!
!! Enable loose uRPF on core interfaces
interface TenGigE0/0/0/1
description CORE-FACING
ipv6 verify unicast source reachable-via any
!
Rate Limiting Specifics¶
| Traffic Type | Recommended Rate Limit | Rationale |
|---|---|---|
| ICMPv6 to SRv6 SID prefixes | 100-500 pps per interface | Prevents topology discovery via traceroute/ping while allowing legitimate OAM |
| SRH packets to control plane (punt path) | 1000-5000 pps | Prevents CPU exhaustion from SRH processing; legitimate SRH traffic is forwarded in hardware |
| BGP updates for SRv6 SID routes | Standard BGP rate limiting | Prevents BGP flooding; use existing BGP protection mechanisms (GTSM, max-prefix) |
| ICMPv6 Packet Too Big for SRv6 | 50-200 pps per interface | Needed for PMTUD but can be abused for recon; rate-limit conservatively |
!! CoPP policy for SRv6-related traffic
policy-map COPP-SRV6
class ICMPV6-TO-SID-BLOCK
police rate 500 pps
conform-action transmit
exceed-action drop
!
class SRH-TO-CONTROL-PLANE
police rate 2000 pps
conform-action transmit
exceed-action drop
!
!
Implementation Vulnerabilities¶
Protocol-level protections do not help if the SRH-processing code itself crashes. Keep SRv6 data planes patched like any other exposed parser.
| CVE | Component | Issue | Fixed in |
|---|---|---|---|
| CVE-2026-23442 | Linux kernel IPv6 / SRv6 (net/ipv6/exthdrs.c, net/ipv6/seg6_hmac.c) | ipv6_srh_rcv() and seg6_hmac_validate_skb() used the per-device IPv6 state returned by __in6_dev_get() without a NULL check. It is NULL when the interface has no IPv6 configuration (e.g. MTU below the IPv6 minimum, or during unregistration), so processing an SRH on such a device could dereference NULL and crash the kernel. NVD scores it 5.5 (Medium, availability impact). Affected since SRv6 landed in 4.10. | Mainline 7.0; stable 5.10.258, 5.15.209, 6.1.175, 6.6.136, 6.12.83, 6.18.25, 6.19.10 |
Hosts and appliances count too
Linux SRv6 is used on hosts, CPEs, virtual routers and SONiC/whitebox control planes (Host-Based SRv6). Track kernel stable updates for every such node, not just routers. Note that for this CVE the dereference happens while the kernel reads the per-interface SRv6 setting itself, so disabling seg6_enabled is not a substitute for patching.
Comparison with MPLS Security¶
SRv6 and MPLS share a common security foundation (the trusted domain model) but differ in important ways. Understanding these differences helps operators transitioning from MPLS to SRv6.
| Aspect | MPLS | SRv6 |
|---|---|---|
| Trust model | Trusted domain -- labels are only meaningful within the operator's network | Trusted domain -- SIDs are only meaningful within the SRv6 domain |
| Label/SID visibility | Labels are locally significant integers (20-bit); not routable outside the network | SIDs are globally routable 128-bit IPv6 addresses; reachable if locator prefix leaks |
| Per-packet authentication | None. MPLS has no mechanism equivalent to HMAC. Label integrity relies entirely on physical and logical network security. | Optional HMAC TLV (RFC 8754) provides per-packet cryptographic authentication of the segment list |
| Boundary filtering | Implicit -- MPLS labels are not present in IP packets from external sources (LDP/RSVP only assign labels internally) | Explicit filtering required -- SRH is a standard IPv6 extension header that any host can generate |
| Topology exposure | Low -- label values reveal nothing about topology; label-to-path mapping requires internal access | Higher -- SRv6 SIDs encode locator information and can be probed via standard IPv6 tools |
| Attack surface | Smaller -- requires access to the MPLS data plane (typically Layer 2 adjacency) | Larger -- any IPv6-capable host can craft packets with SRH targeting the SRv6 domain |
| Encryption | MACsec on links; no native payload encryption | MACsec on links; IPsec integration possible (but complex with SRH) |
The HMAC advantage
While SRv6 has a larger attack surface than MPLS (due to SIDs being routable IPv6 addresses), it also has a stronger optional defense: HMAC provides cryptographic proof that the segment list was created by an authorized node. MPLS has no equivalent mechanism -- label integrity is assumed, never verified.
Migration consideration
Operators moving from MPLS to SRv6 must recognize that the "implicit security" of MPLS (labels not being externally routable) is replaced by explicit security in SRv6 (mandatory boundary filtering). This is the most important operational difference: SRv6 security requires active configuration, while MPLS security was largely passive.
Post-Quantum Transport Encryption¶
SRv6 provides integrity mechanisms (HMAC) but no confidentiality — payload encryption is always a separate transport layer (MACsec on links, IPsec over untrusted transits). That layer now faces its own deadline: adversaries running Harvest Now, Decrypt Later campaigns archive encrypted traffic today to decrypt it once quantum computers can break classical key exchange (RSA/DH/ECDH). For data that stays sensitive for a decade or more, classical IPsec is already insufficient.
The first validated design pairing SRv6 with post-quantum cryptography is Cisco's 2026 mission-critical CVD, which runs the SRv6 fabric over:
- WAN MACsec authenticated via EAP-TLS with ML-KEM (FIPS 203) key exchange on private links
- GRE over IPsec with IKEv2 hybrid key exchange (RFC 9370: classical ECDH + ML-KEM-1024) on untrusted transits
- IS-IS link affinities tagging encrypted vs. unencrypted interfaces, so Flex-Algo and SR-TE policies can treat encryption as a routing constraint
See Mission-Critical Networks for the full architecture, including the IKEv2 fragmentation gotcha that comes with ML-KEM's large keys.
Further Reading¶
- SRH Mechanics & Packet Walk - SRH format including HMAC TLV
- OAM & Troubleshooting - Detecting SRv6 issues
- Mission-Critical Networks - Quantum-safe SRv6 fabric design (PQC MACsec/IPsec)
- RFC 8754 - SRH specification including HMAC
- RFC 9602 - The 5f00::/16 SID block, useful for boundary filtering
References¶
- RFC 8754 - IPv6 Segment Routing Header - Section 7 defines HMAC TLV and security considerations for SRH
- RFC 8402 - Segment Routing Architecture - Section 8 covers SR security properties and the trusted domain model
- draft-ietf-spring-srv6-security - IETF draft on security considerations specific to SRv6 deployments
- Cisco CVD: Quantum-Safe SRv6 Fabric for Mission-Critical Networks - Validated design running SRv6 over post-quantum MACsec and IPsec transport
- CVE-2026-23442 (CVE.org) - Linux kernel CNA record: NULL idev dereference in SRv6 SRH and HMAC paths, with fixed stable versions
- CVE-2026-23442 (NVD) - NVD entry with CVSS 3.1 score (5.5, CWE-476)