DHCP is convenient because clients can join a network without a technician entering an IP address, subnet mask, default gateway and DNS settings on every device. That convenience also creates an opportunity for a rogue device to act as a DHCP server and distribute the wrong configuration. DHCP snooping limits that risk at the switch: it distinguishes ports that may carry server responses from ports where clients should ordinarily be requesting leases.
The feature is not a replacement for access controls, a DHCP server or endpoint protection. Its value is narrower and testable: it can reject prohibited DHCP messages and maintain a binding database that other switch protections use. A good deployment protects legitimate address assignment without interrupting authorized DHCP relay and renewal traffic.
Where a rogue server enters the exchange
A client without an address normally sends a DHCPDISCOVER broadcast. A DHCP server or relay path makes an offer available; the client requests the selected address and receives an acknowledgement. The familiar Discover–Offer–Request–Acknowledge sequence is often abbreviated DORA, although renewals and rebinding follow different message paths.
On an unmanaged or unprotected Layer 2 segment, an unauthorized device can answer quickly with a lease that supplies an attacker-controlled gateway or DNS server. The client’s operating system may accept the lease unless another control intervenes. A separate DHCP starvation attack tries to consume server leases or switching resources by generating many requests, sometimes with changing source MAC addresses. The two attacks are related but not identical, and DHCP snooping does not guarantee that every exhaustion tactic will be stopped by one configuration line.
Trusted and untrusted ports are different privileges
An access port facing an ordinary workstation is normally untrusted. A switch permits the client’s legitimate requests but rejects DHCP server messages arriving on that interface. A port toward an authorized DHCP server, or a switch/router carrying legitimate DHCP relay responses, is configured as trusted according to the topology.
The distinction is operational, not moral: a trusted interface is allowed to carry DHCP server responses; it does not confer trust to every host connected beyond it. Blindly trusting a broad downstream trunk can defeat the control if an unauthorized DHCP server sits on one of the trunk’s VLANs. Conversely, leaving the real server path untrusted can cause every new client to fail address assignment.
Trust configuration is also platform- and topology-dependent. A routed relay interface and an L2 access or trunk interface do not necessarily process DHCP packets in the same way. Validate the actual traffic path on the switch model and software release, especially when a relay adds Option 82 information. Cisco’s documented DHCP-snooping behavior across Layer 3 hops illustrates why simple rules about “all Offers are always dropped on untrusted ports” need topology context.
How the binding database is built and used
As the switch observes successful DHCP exchanges, it can record a binding containing the assigned IP address, client MAC address, VLAN, interface, lease duration and binding state. That record establishes which address a particular client may legitimately use on that port. The binding table is not an all-knowing IP inventory: clients with manually configured addresses may not appear unless the platform supports and uses static source bindings.
Bindings matter beyond DHCP. Dynamic ARP Inspection (DAI) can validate certain ARP claims against DHCP-snooping bindings or explicitly configured ARP access lists. IP Source Guard can restrict source IP traffic using the dynamic database and static entries. Those controls have distinct enforcement points. Enabling them before planning for static-IP printers, servers or infrastructure endpoints can break legitimate traffic.
A persistent binding database may help preserve trust information across a restart, but its durability depends on platform support and configuration. Confirm database-agent placement, writes, restart behavior and reconciliation with current leases. Do not assume that an empty binding table is harmless simply because existing hosts can still communicate temporarily.
A sensible VLAN rollout
Choose one pilot access VLAN and document the actual DHCP-server and relay path. Identify client-facing interfaces and upstream trusted connections. Enable snooping for the intended VLANs, then apply interface trust deliberately. Configure supported DHCP rate limits on untrusted access ports to reduce processing and resource-exhaustion risk, but do not guess a low threshold that fails under phone, imaging or boot-storm conditions.
Check that legitimate clients receive the intended pool, gateway, DNS servers and lease durations. Use a test client on an untrusted port to verify that an unauthorized server Offer or Acknowledge does not pass. Record switch counters and logs before and after, rather than relying only on a successful connectivity test.
| Observation | Likely area to investigate | Safer first test |
|---|---|---|
| Many clients suddenly stop receiving leases | Real server/relay path is untrusted; VLAN not enabled consistently | Trace the server-return interface and message counters |
| Rogue-server test still succeeds | Trunk or access port incorrectly trusted | Inspect per-interface trust state and VLAN topology |
| Phones or PXE clients fail intermittently | Rate limit, relay handling or Option 82 interaction | Compare packet captures and DHCP logs during startup |
| DAI blocks legitimate static equipment | Binding absent for non-DHCP client | Review static bindings or inspection exceptions |
| Lease assignment works but IP Source Guard fails | Binding did not persist or does not match port/VLAN | Inspect dynamic and static source bindings |
DHCP starvation exhausts available leases and should be diagnosed separately from a rogue DHCP server that advertises an incorrect default gateway.
Option 82 and multi-switch networks
A relay agent or capable switch can add DHCP Option 82 information that identifies where a request entered the network. This can assist policies or address assignment, but it also changes what downstream and upstream equipment must accept. Some designs expect Option 82 from a trusted switch or relay; others will reject relayed packets under a mismatched trust configuration. Cisco has documented different handling for relayed DHCP messages depending on whether they arrive on a routed, trusted Layer 2 or untrusted Layer 2 interface.
When multiple switches are involved, do not assume every DHCP server path stays on the same VLAN and port type. A client on an access switch may traverse an 802.1Q trunk, then reach a DHCP relay on a routed interface. Record this chain before setting trust. If a packet capture shows a relayed message being dropped, inspect the relay address, Option 82 contents and software-specific validation behavior, not only the destination server’s logs.
Verification matters more than a green configuration
After enabling the feature, collect the intended VLAN list, trusted port list, untrusted rate limits, snooping counters, dynamic bindings and test results. Verify that authorized DHCP leases renew successfully, not merely that a single client obtained one new lease. Renewals may use unicast messages and can expose different path assumptions from initial broadcast discovery.
Monitor for unexpected new trusted interfaces, changes to the server/relay topology and sustained DHCP drops. Where a network has redundant paths, repeat the test after a failover so the secondary path does not silently bypass snooping or reject all leases. Treat high drop counts as a diagnostic trigger: they can indicate a real attack, a legitimate onboarding event or a misconfigured uplink.
DHCP snooping is most effective as a precise port-and-VLAN enforcement control within a documented switch design. It narrows who can supply leases, helps build address-to-port evidence, and supports DAI or source guard. Its success depends on correct trust placement and a maintained exception process—not on assuming every DHCP packet from an unexpected interface is malicious.