Attack Analysis

A UDP flood rarely announces itself with a clean outage. It starts as a bandwidth climb, a few odd packet counters, and a firewall whose state table suddenly looks busy. Knowing which data to watch, and in what order, decides whether you catch it in seconds or minutes.
UDP flood
A UDP flood is a volumetric attack that sends a high rate of UDP datagrams to a target, exhausting bandwidth, CPU or connection-tracking capacity. Because UDP is connectionless, there is no handshake to validate the sender, which makes the traffic cheap to generate and hard to filter naively.
The signals that matter most fall into a few groups. Track them continuously, not only during an incident, so you have a baseline to compare against.
- Packets per second (pps): often spikes before bandwidth does, especially with small payloads.
- Bits per second (bps): shows link saturation risk.
- Top destination ports: a sudden concentration on one port, or random sweeps across many ports, is a strong indicator.
- Packet size distribution: floods often use uniform sizes, either tiny packets or near-MTU payloads.
- Source IP diversity: an abnormal count of unique sources, or a single source at huge volume.
- ICMP "port unreachable" rate: your host replying to closed ports is a classic side effect.
- Device health: CPU, interrupt load, conntrack/state table usage and interface drops.
Tip: Alert on pps and on the ratio of UDP to total traffic, not only on bps. A small-packet flood can cripple a router while the bandwidth graph still looks moderate.
UDP attack types
UDP attacks differ mainly in how traffic reaches the victim and what it targets. Identifying the type from your telemetry tells you which mitigation to apply.
| Type | What you will see | Key data to check |
|---|---|---|
| Direct UDP flood | High pps to one or many ports | Destination port spread, source count |
| Reflection/amplification (DNS, NTP, memcached, etc.) | Large responses from legitimate servers you never queried | Source ports such as 53, 123, 11211; large packet sizes |
| Fragmented UDP | Many fragments, no L4 header in follow-up pieces | Fragment ratio, reassembly errors |
| Application-targeted UDP | Valid-looking packets to a game, VoIP or custom service port | Per-service request rate, payload patterns |
Reflection attacks deserve special attention. Your own servers did not ask for those responses, so inbound UDP from well-known service ports to hosts that do not run resolvers is a reliable red flag.
UDP spoofing
UDP spoofing means forging the source IP address of a datagram. Since no handshake confirms the sender, an attacker can claim any address, hiding their origin and, in reflection attacks, redirecting responses to the victim.
For defenders, spoofing changes how you read source data. A flood from thousands of "different" IPs may be one machine, and blocking sources one by one is futile. Look instead for indirect evidence:
- TTL values that are inconsistent for a single claimed source or network.
- Sources from unallocated or bogon ranges.
- Traffic arriving on an interface where that source could never legitimately appear.
- Uniform payloads and timing across supposedly unrelated sources.
At the network edge, ingress filtering such as BCP 38 (RFC 2827) reduces spoofed traffic, but you cannot rely on every upstream network to apply it. That is why behavior-based filtering matters more than source blocklists.
UDP flood attack prevention
Prevention combines local hardening with upstream filtering. Local rules shrink your exposure, but a volumetric flood that saturates your link must be stopped before it reaches that link.
Local hardening
- Close or firewall every UDP port you do not need, and restrict management services to known addresses.
- Disable open resolvers and unused UDP services so your hosts cannot be abused as reflectors.
- Rate-limit ICMP unreachable responses.
- Apply per-source and per-port rate limits where your hardware supports it.
- Use ingress filtering and reverse-path checks on edge routers.
Upstream mitigation
When attack volume exceeds your uplink, only filtering in front of the link helps. WireShield's filtering covers network- and transport-layer volumetric and protocol attacks, including UDP flood, SYN flood, DNS amplification and fragment attacks. HTTP/L7 and service-specific needs are scoped together during a technical assessment.
| Approach | Best for | Details |
|---|---|---|
| Remote DDoS protection | Hosting providers, agencies, businesses with distributed infrastructure | Traffic is filtered remotely before it reaches your network |
| WireShield Edge (on-site protection) | Enterprise networks and providers needing local inline filtering | Protection deployed at your own location |
UDP flood test
A UDP flood test checks whether your monitoring, alerting and mitigation actually work under load. Done carelessly, it becomes an outage or a legal problem, so planning matters more than the tool.
Warning: Only test infrastructure you own or have written authorization to test. Notify your upstream provider first, since unannounced high-volume traffic can trigger their abuse systems or affect other customers.
A sensible test plan looks like this:
- Define the goal: validate alert thresholds, failover or filtering response time.
- Start in a maintenance window with a low rate and increase in steps.
- Record pps, bps, CPU, drops and time-to-detect at each step.
- Test several packet sizes, since small and large payloads stress different components.
- Review the results and adjust thresholds, then repeat.
Keep in mind that a small lab test does not reproduce a large internet-scale attack. Its value lies in confirming that detection works and that your team knows the response procedure.
Frequently asked questions
What is the first metric to check during a suspected UDP flood?
Check packets per second on the affected interface, then compare UDP share against your normal baseline. A sharp pps rise with unchanged business activity is the quickest indicator. Follow it with destination port distribution and device CPU to see whether the load is overwhelming forwarding or only consuming bandwidth.
Can a UDP flood happen without high bandwidth?
Yes. Floods of tiny packets can exhaust router or firewall processing capacity while bandwidth stays modest. This is why packet rate and device load need alerts of their own, rather than relying on throughput graphs alone.
How do I tell reflection traffic from a direct flood?
Reflection traffic arrives from legitimate service ports such as 53 or 123, usually with large responses to requests you never sent. Direct floods tend to target destination ports with uniform or random packets from spoofed or botnet sources.
Next steps
Good UDP flood defense starts with a baseline: know your normal pps, port mix and source diversity, and every deviation becomes easier to read. If you want filtering in front of your network, review remote DDoS protection or WireShield Edge, and ask for a technical assessment to define the scope for your setup.
Related content
- SYN flood detection: which metrics reveal half-open connection abuse
- DNS amplification attacks and how to avoid becoming a reflector
- Fragment attacks: how fragmented packets bypass simple filters
- Building a DDoS runbook for hosting providers
- Remote vs. on-site DDoS protection: choosing the right model