Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

What Are Reflection and Amplification Attacks?

Reflection hides the attacker behind innocent third-party servers; amplification makes those servers return far more data than they received. Together they produce the largest volumetric DDoS attacks.

Illustration for the cybersecurity term Reflection and Amplification Attacks
Home / Blog / Cybersecurity Terms / What Are Reflection and Amplification Attacks?
17 November 2026

What Are Reflection and Amplification Attacks?

What Are Reflection and Amplification Attacks?

Reflection and amplification are two distinct ideas that are almost always used together, which is why they are usually named as a pair. Reflection means bouncing traffic off an innocent third party so that the victim receives it from that party’s address rather than the attacker’s. Amplification means choosing a service whose reply is substantially larger than the request that triggered it. Combined, they let a modest amount of attacker bandwidth deliver an overwhelming flood.

The Mechanism

The technique depends on UDP, which performs no handshake. A UDP server receives a request and replies to whatever source address the packet claims—there is no verification that the sender is really there.

So the attacker sends a small request to a public UDP service with the source address forged to the victim’s. The service replies, to the victim. Repeat across hundreds of thousands of such services and the victim is buried under traffic from legitimate, identifiable servers that have done nothing wrong and have no idea they participated.

Three consequences follow. The victim cannot block the attacker, because the attacker’s address never appears. Blocking the reflectors means blocking real infrastructure, including DNS resolvers the victim may need. And the attacker’s own bandwidth requirement is divided by the amplification factor.

Which Protocols Amplify

ProtocolApproximate factorNote
DNS28–54xOpen resolvers; a short query returns a large record set
NTP (monlist)Up to 556xA debugging command returning the last 600 clients
SSDP~30xConsumer routers and media devices exposed to the internet
CLDAP56–70xMisconfigured directory services on UDP 389
MemcachedUp to 51,000xCaching servers bound to a public interface with no authentication
WS-Discovery300–500xNetwork cameras and printers
Mitel MiCollab (TP240)Extreme, up to 4.3 billion to 1CVE-2022-26143, a test command producing sustained output

The memcached campaign of February 2018 produced the defining example: a 1.35 Tbps attack against GitHub, followed days later by a 1.7 Tbps attack reported by NETSCOUT. Roughly 50,000 exposed memcached servers were enough, because each 15-byte request returned up to 750 kilobytes.

Earlier, NTP amplification dominated 2014 before the monlist command was disabled by default, and DNS amplification has never gone away because open resolvers keep reappearing.

Why It Persists

The root cause is that most networks do not filter outbound traffic by source address. BCP 38, published as RFC 2827 in 2000, specifies that a provider should drop packets leaving its network with a source address it has not assigned. If every network implemented it, source spoofing would be impossible and the entire technique would collapse.

Adoption remains incomplete after a quarter of a century, because the cost falls on the network that would deploy it while the benefit falls on everyone else. The MANRS initiative encourages operators to commit to filtering, and many UK providers participate, but the long tail of smaller networks worldwide keeps the capability alive.

Defending Your Own Services

There are two separate responsibilities here.

Do not be a reflector. Audit what you expose on UDP. Memcached, Redis, and similar services should never be reachable from the internet. DNS resolvers should be restricted to your own clients rather than open to the world, and authoritative servers should implement response rate limiting. Disable NTP monlist, block SSDP and WS-Discovery at the perimeter, and scan your own address space periodically with the tooling the Shadowserver Foundation publishes.

Survive being a target. Volumetric attacks saturate your upstream link before any on-premises equipment sees them, so mitigation must happen upstream. That means a scrubbing provider or an always-on CDN, an agreed activation procedure your team has actually rehearsed, and a provider contact who answers out of hours. Keep DNS time-to-live values low enough to redirect traffic quickly, and know in advance which services must stay available.

Application-layer techniques such as HTTP/2 Rapid Reset exhaust server resources instead of bandwidth and need different controls, so a complete DDoS plan covers both. Our DoS/DDoS Testing Services measure where your capacity actually ends and whether the mitigation path works when called upon.

FAQ – Frequently Asked Questions

What Is the Difference Between Reflection and Amplification?

Reflection conceals the attacker by bouncing traffic off a third-party server, so the victim sees the reflector’s address. Amplification increases the volume, because the reflector’s reply is larger than the request that produced it. An attack can reflect without amplifying, but the two are combined in practice because amplification is what makes a small botnet capable of saturating a large link.

Why Can These Attacks Not Simply Be Blocked by Source Address?

Because the source addresses are genuine servers, not attackers. Blocking them means blocking real infrastructure—public DNS resolvers, time servers, and content providers—that your users and systems may legitimately need. The traffic also arrives at full volume before any filtering device you control, so blocking at the perimeter does not relieve the saturated upstream link.

How Do We Know Whether We Are Being Used as a Reflector?

Look for high volumes of outbound UDP responses to addresses that never initiated a connection, particularly on ports 53, 123, 389, 1900, 3702, and 11211. The Shadowserver Foundation publishes free daily reports of exposed and abusable services by network, and subscribing to their feed for your address space is one of the simplest and most useful things a network team can do.

Would BCP 38 Really Stop These Attacks?

Universal deployment would, because source address spoofing is the foundation of the technique and ingress filtering prevents it. The obstacle is economic rather than technical: a network that implements filtering protects other people’s networks, not its own, so adoption has been slow since publication in 2000. The MANRS initiative exists to make participation visible and expected.

Related Terms

You May Be Interested In These

What Is Recovery in Incident Response?

Cybersecurity Terms

17 November 2026

What Is Recovery in Incident Response?

Read More
What Is Nuclei?

Cybersecurity Terms

17 November 2026

What Is Nuclei?

Read More
What Is SASE?

Cybersecurity Terms

17 November 2026

What Is SASE?

Read More