Address Resolution Protocol (ARP) answers a narrow but essential question in an IPv4 Ethernet network: which link-layer address should a host use to deliver a frame to the next IPv4 hop? A host might know the destination IP address and have a correct route, yet still need a neighbor’s MAC address before it can transmit an Ethernet frame. ARP is the local link between those decisions. It is neither an Internet-wide directory nor a substitute for routing.
The distinction matters because older explanations often describe ARP as universal IP-to-MAC resolution. ARP resolves IPv4 neighbors on a local link; IPv6 uses Neighbor Discovery over ICMPv6 instead. Understanding the boundary prevents misleading troubleshooting, inaccurate security advice, and failed attempts to diagnose an IPv6 problem using an IPv4 ARP table.
What ARP actually resolves
Consider an IPv4 host at 192.0.2.10/24 sending traffic to 192.0.2.20. Its routing table identifies the destination as directly connected. Before creating an Ethernet frame, it checks a neighbor cache for the MAC address associated with 192.0.2.20. If there is no usable entry, it broadcasts an ARP request on the local Layer 2 segment. The owner of that IPv4 address replies with its MAC address, and the sender can transmit the frame. The IP destination stays 192.0.2.20; the Ethernet destination is the discovered MAC.
Now change the destination to 198.51.100.20, outside that /24. The host does not broadcast an ARP query for the remote machine. It routes toward its configured next hop, such as 192.0.2.1, and resolves the next-hop MAC on the local link. Routers build a new link-layer frame at each routed hop. This explains why a working ARP entry for the default gateway does not prove that a remote prefix is reachable.
Requests, replies, caches, and the limits of a broadcast
An ARP request asks who has a specified IPv4 address; it is ordinarily broadcast within the sender’s VLAN. The owner responds with a reply that supplies a sender protocol address and sender hardware address. Neighbor-cache entries allow subsequent packets to avoid repeating the exchange for every frame. Cache aging and reachability behavior vary by operating system and device implementation, so an engineer should not assume one universal ARP timeout.
ARP broadcasts stay inside their Layer 2 broadcast domain. A VLAN boundary normally prevents an ordinary request from reaching hosts in another VLAN; inter-VLAN communication depends on a router or a Layer 3 switch. Proxy ARP is a separate, explicitly configured behavior that can make a router answer for another IPv4 host in some designs. It should not be assumed present or used to paper over an incorrect mask or route.
ARP is not routing, DNS, or IPv6 Neighbor Discovery
DNS maps names to address records, routing selects an IPv4 next hop, and ARP maps that local next-hop IPv4 address to link-layer delivery information. Those functions are connected but independent. A successful DNS lookup does not guarantee a route, and a correct route does not guarantee the local gateway’s MAC can be resolved. Troubleshooting works best when these decisions are tested separately rather than calling each failure an “ARP issue.”
For IPv6, Neighbor Solicitation and Neighbor Advertisement perform address resolution within Neighbor Discovery. They use ICMPv6 and solicited-node multicast rather than the ordinary IPv4 ARP broadcast exchange. Router Advertisements, Duplicate Address Detection, and neighbor reachability add functions that ARP itself does not provide. The relevant troubleshooting path is an IPv6 neighbor table and the appropriate ICMPv6 control traffic, not an assumption that ARP crosses both IP versions.
Gratuitous ARP, address moves, and duplicate-address risks
A host can announce an IPv4-to-MAC binding without first being asked, commonly called gratuitous ARP. It is useful after a virtual IP or host interface moves to a different device because neighbors may have a stale mapping. ARP announcements also help reveal duplicate IPv4 address assignments in some operating environments. The exact handling depends on the receiving systems, ARP settings, and whether an intermediary device caches the announcement.
That flexibility also permits abuse: ARP replies are not cryptographically authenticated by default. A malicious or misconfigured node may convince peers to associate the gateway IPv4 address with the wrong MAC and redirect or interrupt local traffic. A source MAC displayed in a packet capture is evidence about a claimed binding, not proof that the sender was authorized to make it.
Dynamic ARP Inspection and trustworthy bindings
On supported Cisco Catalyst switches, Dynamic ARP Inspection (DAI) can validate ARP packets received on untrusted ports against IP-to-MAC bindings learned through DHCP snooping. Cisco documents that an invalid packet is discarded. This protection requires correct VLAN configuration, trusted uplink selection, and a binding source; a switch cannot reliably check an unmanaged static-address host against a DHCP binding that does not exist. For static hosts, approved ARP access lists or platform-specific binding mechanisms may be required.
A trust-state mistake can create an outage just as surely as an attack: trust an untrusted edge and spoofing may pass, or mark a transit port untrusted without supplying the necessary bindings and legitimate hosts can lose connectivity. Verify actual bindings, the origin of DHCP offers, ARP inspection drops and port trust before enabling DAI widely. Cisco’s documentation about Dynamic ARP Inspection makes these prerequisites explicit; it is a control in a layered design, not an automatic guarantee against every man-in-the-middle technique.
A practical failure-isolation sequence
Start with the reported destination. Confirm the host IPv4 address, subnet mask, default gateway and effective route. If the destination is local, examine the IPv4 neighbor cache and packet capture on the correct VLAN. If remote, test neighbor resolution for the selected next hop before diagnosing any further path. Look for unanswered ARP requests, replies carrying unexpected MAC addresses, duplicate addresses, access-port VLAN errors or MAC moves in switch telemetry.
Next correlate the neighbor entry with switch MAC-address learning. A known MAC on the wrong interface can indicate cabling, a moving virtual machine, a loop or a security incident. An empty neighbor cache immediately after an interface restart can be normal. Compare before-and-after captures under an isolated, approved test rather than clearing all production neighbor entries or disabling security features reflexively. This layered workflow complements broader network troubleshooting because the order of the failure is more useful than a single symptom.
How to validate an ARP explanation in a lab
Use two test hosts and one router or multilayer switch with documented interface/VLAN addresses. Capture an initial exchange between two IPv4 hosts in the same subnet, then a flow to a remote test subnet. Record which IP address appears as the ARP target in each capture. Observe the cache, switch MAC table and default-route selection. Repeat after moving the remote service to another subnet; the host should still resolve its local next hop, not the remote machine.
For a security-focused lab, isolate a disposable VLAN and verify whether a supported switch has DHCP snooping bindings before trying DAI. Confirm that authentic hosts work with untrusted access ports and that intentionally invalid test bindings are rejected. Keep the exercise separate from production devices. Then repeat with IPv6, substituting Neighbor Solicitation/Advertisement and the IPv6 neighbor table. That contrast makes the important version boundary visible.
Operational evidence that establishes an ARP diagnosis
A suspicious ARP cache entry is a clue, not a verdict. Record the exact VLAN, client IP, default-gateway IP and MAC addresses observed from the endpoint. On a managed switch, correlate that MAC address with the current forwarding table and interface, then compare the binding with DHCP snooping records and the authorized gateway hardware or first-hop redundancy virtual MAC. A virtual gateway can legitimately advertise the same IP while its active device changes. Separate an expected failover update from an unauthorized responder by checking the timing of gateway-state transitions and whether the MAC matches the configured redundancy protocol.
Capture both ARP requests and replies during the incident. Determine whether conflicting replies arrive for one address, whether the purported sender lives in the correct VLAN, and whether a recent switch-port move can explain the change. Authentication of an application session with TLS may protect its contents even if an attacker redirects the traffic, but ARP poisoning can still cause denial of service and exploit clients that ignore certificate warnings. Treat network enforcement, endpoint authentication and incident logging as complementary protections, not interchangeable ARP fixes.
If the source of the binding is a server with a static address, verify any ARP inspection allowlist against approved address documentation before imposing a drop policy. DHCP snooping tables cannot describe every legitimate statically addressed host. In a production incident, first collect a short evidence window, then make one targeted VLAN/interface control change and monitor ARP inspection counters and user reachability. A broad trust-state change without a rollback record can hide spoofing or disrupt an entire access network.
IPv6 migration, device behavior and realistic control limits
ARP remains relevant wherever IPv4 operates, including dual-stack networks. Enabling IPv6 does not remove the attack surface of IPv4 neighbors or silently convert ARP into Neighbor Discovery. IPv6 neighbor advertisements, router advertisements and duplicate-address detection have separate protections and failure modes; where supported, Cisco IPv6 First Hop Security controls should be configured according to the exact platform, VLAN design and release. Operational dashboards must distinguish an IPv4 ARP-table issue from an IPv6 neighbor-cache issue.
Static ARP entries may be appropriate for a tiny, stable test environment but are burdensome for roaming clients, DHCP pools and high-availability gateway failover. Similarly, endpoint ARP-monitoring tools can alert on changes, yet an alert alone does not enforce source authenticity. Prioritize secure access-port design, DHCP-snooping binding integrity, DAI where applicable, authenticated application traffic, and a documented recovery process. Future automation should validate approved source-of-truth bindings, not claim that AI or blockchains make ARP inherently trustworthy.
The architectural takeaway
ARP is indispensable inside many IPv4 Ethernet networks precisely because it is limited. It resolves the next local link-layer destination; it does not discover remote hosts, determine the best route, validate identity or protect an entire routed network. A useful explanation should identify the local subnet boundary, the next-hop IP, the relevant cache and the packet actually seen on the wire.
The same clarity helps design segmentation. A routed interface creates a boundary for ordinary ARP broadcasts, but separate subnets do not automatically provide access control. Apply explicit routing and security policy when traffic between networks needs restriction. For related address-planning decisions, see the practical role of CIDR prefixes and the effects of VLAN layout. Treat source validation, neighbor discovery and reachability as distinct checks rather than as one catch-all networking feature.