Reconnaissance, scanning and address privacy
Explain why brute-force /64 sweeps fail and which reconnaissance techniques still find IPv6 hosts. Scan an IPv6 lab with capable tools and interpret the results. Recommend interface-identifier strategies by host role (RFC 8064, RFC 7217).
Take the quiz · pass mark 70% · 9 questions
Explanation
IPv4 reconnaissance is simple: sweep the /24, which is 256 addresses, in a moment. A /64 has 2^64 ≈ 1.8 × 10^19 addresses. Scanning one at a million probes per second would take over half a million years. So brute-force sweeping is dead — but that does not mean hosts are hidden. Attackers changed technique, and RFC 7707 catalogues how.
Why sweeps fail, and what replaces them
Reconnaissance moves from brute force to finding patterns and using other sources:
- Predictable interface identifiers. If hosts use patterned IIDs, the search space collapses:
- Low-byte addresses (
::1,::2,::25): only a few hundred to probe. Common on servers and routers. - EUI-64 IIDs from the MAC (
...ff:fe...in the middle, L1): reveal the vendor OUI and shrink the search to that vendor's range. - IPv4-based IIDs (the host's IPv4, e.g.
192.0.2.25, encoded in the last 32 bits): try the known IPv4 addresses. - Service-port or "wordy" IIDs (
::80,::443,::dead:beef): a small favourites list.
- Low-byte addresses (
- DNS. Forward DNS enumeration (names you can guess or get from certificate transparency logs), and reverse DNS of a /64 (walking
ip6.arpa, or noticing a server that returns NXDOMAIN vs NOERROR). Public certificate transparency logs list hostnames that then resolve to AAAA records. - Local multicast and ND. On-link,
ff02::1(all nodes) and listening to ND/MDNS reveal every host instantly — no guessing. This is why on-link recon is trivial and perimeter obscurity is weak. - Leaked addresses. Server logs, email headers (
Received:), application traffic, peer-to-peer and gaming all expose client addresses. - Routing and scanning aids. Traceroute reveals router addresses (often low-byte); tools like
scan6/THC-IPv6 andnmapgenerate candidate lists from these patterns rather than sweeping.
So the defence is not predictability through obscurity but reducing patterns and controlling the other sources.
Interface-identifier strategy (RFC 8064, RFC 7217)
RFC 7721 analyses the privacy and security trade-offs of each IID type. RFC 8064 sets the modern default: stable, opaque IIDs generated per RFC 7217, not derived from the MAC.
- RFC 7217 stable-privacy IIDs: a hash of the prefix, interface, a secret key and a counter. The address is stable per subnet (good for server reachability, ACLs and logging) but opaque (no MAC, no vendor, no pattern) and different on each network (does not track the device across networks). This is now the SLAAC default on major operating systems and is the recommendation for most hosts.
- Temporary addresses (RFC 8981, L3.04): short-lived random addresses used as the source for outbound connections, so clients are not tracked over time by their address. Used alongside a stable address, not instead of it.
- EUI-64: avoid. It leaks the MAC (vendor, and device tracking across networks) and is predictable. RFC 8064 recommends against it.
- Manual low-byte (
::25): fine and convenient for servers and routers whose addresses are published in DNS anyway — there is no privacy to protect on a public server, and operational clarity wins. Do not use low-byte for client hosts.
Recommendation by role
| Host role | IID strategy | Why |
|---|---|---|
| Client (laptop, phone, workstation) | RFC 7217 stable + RFC 8981 temporary | Privacy and anti-tracking; no inbound to protect |
| Published server | Manual low-byte or memorable, in DNS | Address is public anyway; clarity and stable ACLs |
| Router/infrastructure | Manual low-byte | Operability; already known |
| IoT / appliances | RFC 7217 if possible; never EUI-64 | Avoid MAC leakage and tracking |
Privacy is not the only goal: on a server you want a stable, known address. Match the IID to the role.
What this means for defence
- Don't assume a big subnet hides hosts; assume an attacker can enumerate via DNS, logs and patterns.
- Reduce patterns on clients (stable-privacy + temporary), but keep servers deliberately findable and firewalled (L5.05).
- Protect the real recon channels: lock down reverse-DNS zone transfers, watch certificate transparency for your domains, and apply first-hop security so on-link recon is at least logged.
- Monitor for scanning (bursts of NS for unused addresses tie back to neighbour-cache exhaustion, L5.03).
ictNetworks scenario
ictNetworks configures customer-managed clients for RFC 7217 stable addresses plus temporary addresses, and gives servers memorable low-byte addresses (::53 resolvers, ::25 mail, ::80/::443 web) that are public in DNS and tightly firewalled. It does not rely on address secrecy for security. Its SOC watches for reverse-DNS walking of customer /64s and for NS floods to unused space, and it treats a server found on EUI-64 as a misconfiguration to fix.
Drawing diagram…
Worked examples
Example 1: How long is a /64 sweep?
A /64 holds 2^64 addresses. At 1,000,000 probes per second that is 2^64 ÷ 10^6 ≈ 1.8 × 10^13 seconds ≈ 585,000 years. Brute force is not an option.
Example 2: How patterns collapse it
If a site uses only low-byte addresses ::1 to ::ffff, the attacker probes at most 2^16 = 65,536 addresses — seconds, not millennia. An EUI-64 address reveals the 24-bit OUI, reducing a vendor-specific search to the remaining 24 bits (2^24), still large but far from 2^64.
Example 3: Spotting an EUI-64 address
2001:db8:3000:10:0250:56ff:fe12:3456 has ff:fe in the middle of the IID and the 7th bit of the first byte flipped: it is an EUI-64 IID from MAC 00:50:56:12:34:56 (OUI 00:50:56 = VMware). An attacker learns the host is a VMware VM and can target that OUI range. RFC 8064 says to avoid this.
Lab
Goal: Discover hosts in your own lab, then compare predictable IIDs with stable-privacy IIDs.
You need: a lab /64 with several hosts configured different ways: some low-byte (::1, ::25), one EUI-64, one RFC 7217 stable-privacy, one with temporary addresses; the lab DNS with some AAAA records and a reverse zone; nmap and THC-IPv6 (scan6, dnsrevenum6) on an attacker host; a router for off-link tests.
- Off-link brute force fails: from off-link, try
nmap -6across a range of the /64 and show it finds nothing in reasonable time. Explain the arithmetic. - On-link is trivial: from a host on the link, ping
ff02::1and read the neighbour table; show every host appears at once. Explain why perimeter obscurity does not help on-link. - Pattern-based recon: use
scan6to generate and probe low-byte and IPv4-based candidates, and show it finds the low-byte and EUI-64 hosts but not the RFC 7217 one. Identify the EUI-64 host's vendor from its IID. - DNS recon: enumerate AAAA records by guessing names and by walking the reverse zone (
dnsrevenum6); show which hosts you find via DNS regardless of their IID. - Privacy comparison: on the RFC 7217 host, show the address is stable across reboots but different on another subnet; on the temporary-address host, show outbound connections use a changing source while a stable address remains for inbound.
- Recommend an IID strategy for each host role in the lab and justify it.
What to submit: the scan outputs, which hosts each method found, the EUI-64 vendor identification, the stability/temporary evidence, and your per-role recommendations.
Summary
- Brute-forcing a /64 is infeasible (~585,000 years at 1M/s); recon uses patterns, DNS, logs and on-link multicast instead.
- Predictable IIDs (low-byte, EUI-64, IPv4-based) collapse the search space; EUI-64 also leaks the MAC and vendor.
- RFC 8064 recommends RFC 7217 stable-opaque IIDs for SLAAC, plus RFC 8981 temporary addresses for client privacy.
- Match the IID to the role: stable-privacy + temporary for clients, memorable low-byte for published servers and routers.
- Don't rely on address secrecy; firewall servers and monitor the real recon channels.
- Next lesson (L5.07): IPsec and VPNs over IPv6.
Quiz imported · 9 questions · pass mark 70%. Lab questions are not auto-marked.