How do IPv6 SLAAC and EUI-64 addressing actually work?
AdminKit explains how IPv6 SLAAC and EUI-64 addressing work together: SLAAC lets a device build its own routable address from a router's advertised /64 prefix plus a 64-bit interface identifier, and EUI-64 is one way to build that identifier — directly from the device's MAC address, which can expose its hardware identity across networks.
By AdminKit team, Netbase JSC · Updated
Key takeaways
- SLAAC (RFC 4862) lets a device build its own IPv6 address: the router advertises a
/64prefix, and the device appends a 64-bit interface identifier it creates itself. No server hands out or tracks the address. - Modified EUI-64 is one way to make that identifier: split the 48-bit MAC, insert
ff:fein the middle, and flip one bit. The result is the same on every network, so it can be used to track a device. - Two replacements exist. Temporary addresses (RFC 8981, which replaced RFC 4941) rotate over time. Stable-privacy addresses (RFC 7217) stay fixed on one network but differ between networks.
ff:fein the middle of the last 64 bits is a strong hint of EUI-64. Check the device’s own settings to be sure, because defaults differ between operating systems and Linux distributions.- SLAAC only assigns addresses. It gives no protection against inbound connections; that is a firewall’s job.
Why did IPv6 need a different way to assign addresses?
Because IPv6 has room for every device to pick its own address, and IPv4 does not. This guide sits in AdminKit’s DNS cluster alongside its other IP- and DNS-addressing guides. IPv4 networks mostly rely on DHCP to hand out addresses, or NAT to let many devices share one public address — workarounds largely necessitated by IPv4’s limited address space. IPv6’s address space is large enough (2^128 addresses) that neither workaround is strictly necessary, which opened the door to a different model: Stateless Address Autoconfiguration (SLAAC), defined in RFC 4862, where a device generates its own complete, globally routable address without asking a server for one.
A router periodically advertises the network’s prefix — almost always a /64, as covered in our guide on subnet and CIDR basics, and typically handed to that router in the first place by its own ISP via IPv6 prefix delegation — and every device on that network independently combines the advertised prefix with a 64-bit identifier it builds itself to form a full 128-bit address. That prefix advertisement is itself one function of Neighbor Discovery Protocol, the broader set of ICMPv6 messages IPv6 uses for everything from router discovery to resolving a neighbor’s link-layer address.
DHCPv6 is the other way a device on an IPv6 network gets configured. Two flags in the router’s advertisement, defined in RFC 4861, tell a host what to do: the M (“managed”) flag says addresses are available through DHCPv6, and the O (“other”) flag says only extra settings, such as DNS servers, are available there, alongside an address the host already built via SLAAC.
What happens when a device joins a SLAAC network?
It gives itself a link-local address first, checks nobody else uses it, then builds a global address from whatever prefix the router advertises. RFC 4862 describes the steps; the table below follows them in order.
| Step | What the device does | Where it is defined |
|---|---|---|
| 1. Link-local address | Combines fe80::/64 with its interface identifier | RFC 4862 §5.3 |
| 2. Duplicate Address Detection | Sends a Neighbor Solicitation for the new address and waits; the address stays “tentative” until no one else claims it | RFC 4862 §5.4 |
| 3. Router discovery | Sends a Router Solicitation, or waits for the next periodic Router Advertisement | RFC 4861 |
| 4. Prefix check | Uses a prefix only if its Autonomous (A) flag is set, and only if prefix length plus identifier length equals 128 bits | RFC 4862 §5.5.3 |
| 5. Global address | Prefix + interface identifier, with the preferred and valid lifetimes copied from the advertisement | RFC 4862 §5.5.3 |
Step 4 is why a /64 matters so much in practice. The interface identifier on Ethernet and Wi-Fi is 64 bits, so RFC 4862 tells a host to ignore a prefix whose length does not leave exactly 64 bits for it. Advertise a /60 or a /72 for autoconfiguration and hosts will not build addresses from it.
The lifetimes in step 5 also explain why an address can show as “deprecated” without being broken. Past its preferred lifetime, an address keeps working for existing connections but is no longer chosen for new ones. Past its valid lifetime, it is removed.
How does EUI-64 turn a MAC address into an interface identifier?
It pads the 48-bit MAC to 64 bits with fffe and flips the universal/local bit. The second half of a SLAAC address — the interface identifier — has to come from somewhere, and one of the original, standardized ways to generate it is the modified EUI-64 format, specified in RFC 4291’s IPv6 addressing architecture (canonical record on the IETF Datatracker). The process is mechanical: take the device’s 48-bit MAC address, split it into two 24-bit halves, insert the fixed 16-bit value fffe between them to pad the result to 64 bits, then flip one specific bit (the universal/local, or “u/l,” bit) to mark the identifier as derived this way. The result is deterministic and reproducible from the MAC address alone — the same NIC always produces the same EUI-64 identifier, on any network.
Here is the full calculation for 00:00:5e:00:53:01, a MAC value that RFC 9542 §2.1.4 reserves for documentation, combined with the IPv6 documentation prefix 2001:db8::/32:
MAC address (EUI-48) 00:00:5e:00:53:01
Split into two 24-bit halves 00:00:5e | 00:53:01
Insert ff:fe between them 00:00:5e:ff:fe:00:53:01
First byte in binary 0000 0000
Flip the u/l bit (value 0x02) 0000 0010 -> 02
Modified EUI-64 identifier 0200:5eff:fe00:5301
Router's prefix 2001:db8:1:2::/64
SLAAC global address 2001:db8:1:2:200:5eff:fe00:5301
Link-local address fe80::200:5eff:fe00:5301
The flip has a reason. In a MAC address, a universal/local bit of 0 means “assigned by the manufacturer”. RFC 4291 Appendix A inverts that bit inside IPv6 identifiers, so that 1 means universal, and RFC 9542 §2.2.1 explains why: it lets administrators type simple local identifiers such as ::1 or ::2 without setting extra bits. The same bit, read by position:
| First byte | Binary | In a MAC address | In a modified EUI-64 identifier |
|---|---|---|---|
00 | 0000 0000 | Universal (vendor-assigned) | Local scope |
02 | 0000 0010 | Local (not vendor-assigned) | Universal scope (derived from a vendor MAC) |
Going backwards is just as easy: drop fffe, flip the same bit, and the original MAC reappears. That is exactly what AdminKit’s IPv6 info tool does in your browser. Paste an address and it splits prefix from interface identifier and, if the identifier is in modified EUI-64 form, shows the embedded MAC; paste a MAC and it shows the identifier that MAC would produce.
This matters because of where the raw material comes from. A MAC address’s first three bytes are its OUI (Organizationally Unique Identifier), assigned by the IEEE Registration Authority to the manufacturer that requested it — the same identifier our MAC OUI glossary entry covers, and one AdminKit’s MAC vendor lookup tool resolves directly. An EUI-64-derived IPv6 address therefore embeds not just a persistent per-device identifier, but one that can often reveal the device’s manufacturer as well, simply by decoding its structure.
Why is a MAC-derived address a privacy problem?
Because the last 64 bits stay identical on every network the device joins. A device’s MAC address doesn’t change, so an EUI-64-derived interface identifier doesn’t change either, regardless of which network the device joins or how many times it reconnects. Combine that constant 64-bit value with whatever /64 prefix a given network advertises, and the resulting address changes its network portion from place to place but keeps the exact same device-identifying tail every time.
That makes it possible, in principle, to recognize the same physical device across completely unrelated networks — home, a coffee shop, a different office — purely from the IPv6 address it generates, without needing cookies, browser fingerprinting, or any cooperation from the device’s software. RFC 8064 states the conclusion plainly: by default, nodes should not use address generation schemes that embed a stable link-layer address in the identifier.
What replaced EUI-64: temporary or stable-privacy addresses?
Both, for different jobs. RFC 4941 defined IPv6 privacy extensions for SLAAC specifically to close this gap: instead of deriving the interface identifier from the MAC address, a device generates a random value, uses it as a temporary address for outgoing connections, and periodically replaces it. RFC 8981 has since obsoleted RFC 4941. Its defaults are a preferred lifetime of one day and a valid lifetime of two days for each temporary address (down from one week in RFC 4941), and it requires fresh temporary addresses when a device moves to a different link.
Temporary addresses cover outgoing traffic, but a device usually keeps a stable address too, for incoming connections and for anything that needs a predictable address. RFC 7217 makes that stable address safe. The identifier is a hash of the prefix, the interface, an optional network ID, a counter for collisions, and a secret key kept on the host. The result stays the same every time the device returns to one network, but looks unrelated on another network. RFC 8064 recommends RFC 7217 as the default way to generate stable SLAAC addresses.
| Method | Defined in | Same identifier on every network? | Changes over time? | Typical use |
|---|---|---|---|---|
| Modified EUI-64 | RFC 4291 App. A | Yes | No, only if the network card changes | Older and embedded systems |
| Stable-privacy | RFC 7217 | No | No, fixed per network | Stable address on hosts that follow RFC 8064 |
| Temporary | RFC 8981 | No | Yes, by default replaced daily | Outgoing connections from clients |
| DHCPv6-assigned | DHCPv6 | Set by the server | Set by the server | Managed networks where M flag is set |
A host that follows these documents typically holds several addresses at once: a link-local, one stable global address, and one or more temporary ones. That is normal, not a misconfiguration.
How can you tell which kind of address a device is using?
Look at the interface identifier, then confirm with the operating system’s own settings. If an address contains ff:fe as the middle two bytes of its last 64 bits, it is very likely modified EUI-64; a random-looking identifier is either stable-privacy or temporary.
Commands to inspect a device (Linux first, then Windows PowerShell):
ip -6 addr show dev eth0
sysctl net.ipv6.conf.eth0.addr_gen_mode net.ipv6.conf.eth0.use_tempaddr
nmcli -g ipv6.addr-gen-mode,ipv6.ip6-privacy connection show "office-lan"
Get-NetIPAddress -AddressFamily IPv6 | Format-Table IPAddress,PrefixOrigin,SuffixOrigin
Replace eth0 with your interface name and office-lan with your NetworkManager connection name. Example output of the first command (illustrative, documentation addresses):
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 state UP
inet6 2001:db8:1:2:9c3e:41d7:a0b2:6f15/64 scope global temporary dynamic
valid_lft 172735sec preferred_lft 86335sec
inet6 2001:db8:1:2:200:5eff:fe00:5301/64 scope global dynamic mngtmpaddr
valid_lft 2591935sec preferred_lft 604735sec
inet6 fe80::200:5eff:fe00:5301/64 scope link
valid_lft forever preferred_lft forever
This host uses EUI-64 for its stable addresses (note ff:fe and the 0200-style identifier) and has temporary addresses turned on for outgoing traffic. On Windows, Microsoft’s Get-NetIPAddress documentation defines SuffixOrigin values: Link means the suffix came from the link-layer address, and Random means it came from a random source.
On Linux, the defaults depend on which layer configures the interface:
| Setting | Values | Default |
|---|---|---|
Kernel addr_gen_mode | 0 EUI-64, 2 or 3 stable-privacy (RFC 7217) | 0 (EUI-64), per the kernel IP sysctl docs |
Kernel use_tempaddr | 0 off, 1 on but prefer stable, 2 on and prefer temporary | 0 for most devices |
NetworkManager ipv6.addr-gen-mode | eui64, stable-privacy, default, default-or-eui64 | Falls back to stable-privacy if no global default is set, per nm-settings-nmcli |
NetworkManager ipv6.ip6-privacy | -1 default, 0 off, 1 prefer public, 2 prefer temporary | Falls back to the kernel’s use_tempaddr |
So two Linux machines on the same network can behave differently, depending on whether NetworkManager, another network manager, or the kernel alone sets up the interface.
AdminKit’s subnet calculator can help you work out the /64 network portion of an address so you can isolate and inspect the 64-bit interface identifier on its own, and the DNS lookup tool lets you check what AAAA (IPv6) addresses a domain or device actually publishes. A published AAAA record with ff:fe in its identifier can reveal the vendor of that server’s network card through its OUI.
Which mistakes come up when reading SLAAC addresses?
| Mistake | Why it is wrong | Better reading |
|---|---|---|
“No ff:fe, so the address is random each day” | Stable-privacy addresses also look random but do not change on the same network | Check for the temporary flag or SuffixOrigin |
“The 02 in the identifier means a randomized MAC” | In an EUI-64 identifier the bit is inverted; 02 means it came from a vendor-assigned MAC | Flip the bit back before reading the MAC |
| “Several global addresses on one interface is a bug” | RFC 8981 hosts keep a stable address plus temporary ones | Expect one stable and one or more temporary per prefix |
| “Any prefix length works for SLAAC” | RFC 4862 ignores prefixes that do not leave a 64-bit identifier | Advertise /64 for autoconfiguration |
| “Every bit in an identifier carries meaning” | RFC 7136 says third parties must treat the identifier as opaque | Use ff:fe as a hint, not proof |
Older IoT devices and embedded hardware are the place to look first: they are more likely to still default to EUI-64 than a current phone or laptop.
Does SLAAC protect a device from inbound connections?
No. SLAAC only handles how an address is assigned — it says nothing about whether that device is protected from unsolicited inbound connections. Because each device gets its own globally routable address rather than sharing one address behind NAT, a router or host firewall has to block unsolicited inbound traffic. That separate concern is covered in IPv6 and home devices: why no NAT means a firewall.
This guide is general information, not a security audit of your specific network — see our disclaimer for what AdminKit's guidance does and does not commit to.
Frequently asked questions
What problem does SLAAC actually solve?
It lets a device get a fully working, globally routable IPv6 address the moment it joins a network, without needing a DHCP server to hand one out and track its lease. Defined in RFC 4862, SLAAC works by having the router advertise the network's /64 prefix, and each device independently combines that prefix with a locally generated 64-bit interface identifier to build its own complete address.
How exactly does EUI-64 turn a MAC address into part of an IPv6 address?
The device's 48-bit MAC address is split in half (24 bits and 24 bits), the fixed 16-bit value fffe is inserted between the two halves to pad it to 64 bits, and one specific bit (the universal/local bit) is flipped to mark the result as a modified EUI-64 identifier, as defined in RFC 4291's addressing architecture. That 64-bit result becomes the last half of the device's IPv6 address, appended to whatever /64 prefix the network advertises.
Why is a MAC-derived IPv6 address considered a privacy problem?
A device's MAC address, and therefore its EUI-64-derived interface identifier, stays constant regardless of which network the device joins. If a device uses EUI-64 addressing on different networks, the same 64-bit identifier appears in its IPv6 address every time, which lets that specific device be recognized and tracked across networks and over time — plus the OUI (vendor prefix) embedded in the MAC often reveals the device's manufacturer. This is exactly the kind of persistent, cross-network identifier that privacy-conscious address generation is designed to avoid.
If EUI-64 has this privacy issue, why do modern devices avoid it by default?
RFC 4941 defined privacy extensions for SLAAC that generate randomized, temporary interface identifiers instead of deriving one from the MAC address, and periodically rotate them; RFC 8981 has since replaced RFC 4941 and shortened the default valid lifetime of a temporary address from one week to two days. Separately, RFC 7217 defines stable addresses that stay the same on one network but differ between networks, and RFC 8064 recommends that scheme over MAC-derived identifiers for stable addresses. Defaults still vary by system: on Linux, for example, the kernel's own defaults are EUI-64 with temporary addresses off, and it is the network manager or distribution configuration that turns on stable-privacy or temporary addresses. Check the device itself rather than assuming.
Does using SLAAC mean a device has no firewall protection?
SLAAC and firewalling are separate concerns, but SLAAC's design does change the threat model: because every device gets its own globally routable IPv6 address rather than sharing one address behind NAT the way IPv4 commonly does, each device is individually reachable from the internet unless something — a router's firewall, a host-based firewall, or both — explicitly blocks unsolicited inbound traffic to it. SLAAC itself doesn't provide that protection; it only handles address assignment.
Sources
- RFC 4291 — IP Version 6 Addressing Architecture (modified EUI-64 format, Appendix A)
- RFC 4862 — IPv6 Stateless Address Autoconfiguration
- RFC 4941 — Privacy Extensions for Stateless Address Autoconfiguration in IPv6
- IEEE Registration Authority — Organizationally Unique Identifiers (OUI assignment)
- IETF Datatracker — RFC 4291, IP Version 6 Addressing Architecture (canonical document record)
- RFC 8981 — Temporary Address Extensions for Stateless Address Autoconfiguration in IPv6 (obsoletes RFC 4941)
- RFC 7217 — A Method for Generating Semantically Opaque Interface Identifiers with SLAAC
- RFC 8064 — Recommendation on Stable IPv6 Interface Identifiers
- RFC 9542 — IEEE 802 parameters, including MAC and EUI-64 documentation values (§2.1.4, §2.2.3)
- Linux kernel documentation — IP sysctl (addr_gen_mode, use_tempaddr)
- NetworkManager — nm-settings-nmcli (ipv6.addr-gen-mode, ipv6.ip6-privacy)
- Microsoft Learn — Get-NetIPAddress (SuffixOrigin)
Want AdminKit to watch this for you?
Free checks tell you the state today. Protective DNS and monitoring tell you when it changes.
Talk to an expert
Huy Nguyen (David), Founder & CEO, Netbase JSC
Ask whether this fix applies to your setup, or get a second pair of eyes on your DNS and mail records.