Quick answer
A smurf attack is a reflection-and-amplification denial-of-service technique from the late 1990s. The attacker sent ICMP echo requests to a network's broadcast address with the victim's address spoofed, so every host replied to the victim at once. A single forged packet became hundreds of replies, amplifying a small upstream into a large flood. It largely died once routers stopped forwarding directed broadcasts by default, per RFC 2644, but its logic lives on in every modern reflection attack.
01What is a smurf attack?
A smurf attack is an amplified denial-of-service attack that turns other people's networks into weapons against a victim. It is named after "smurf", one of the exploit programs attackers used to launch it.
The mechanism has three roles: the attacker, the intermediary network, and the victim. The attacker sends ICMP echo requests (pings, the same packets behind a ping flood) to the broadcast address of the intermediary network, forging the source address to be the victim's. The intermediary's router delivers that broadcast to every host on the segment, and each one dutifully sends an echo reply — to the victim, because that is the source address they see. One request becomes as many replies as there are responding hosts.
CERT described it in advisory CA-1998-01. It is the ICMP ancestor of today's reflection and amplification attacks, differing only in using broadcast domains rather than misconfigured UDP services as the multiplier.
02The amplification maths
The multiplier is simply how many hosts respond to a broadcast. In his original write-up Craig Huegen gave a concrete example: an attacker on a T1 line sending a 768 kbit/s stream of echo requests at a broadcast network of 100 hosts produces 76.8 Mbit/s of traffic at the victim — a hundredfold gain. He observed over 80 Mbit/s generated from roughly 400 kbit/s of input against a 200-host LAN.
That is the entire appeal of amplification: the attacker commits a small, cheap upstream and the intermediaries supply the volume. The same arithmetic, with larger factors, powers the UDP reflectors covered in the amplification attacks guide.
| Quantity | Example value | Note |
|---|---|---|
| Attacker upstream | 768 kbit/s | A single spoofed echo-request stream |
| Responding hosts | 100 | Every live host on the broadcast segment |
| Traffic at victim | 76.8 Mbit/s | Input x responders |
| Amplification factor | ~100x | Set purely by host count on the segment |
03Why smurf largely died
Smurf depended on networks forwarding and answering packets sent to their directed-broadcast address from outside. Remove that behaviour and the multiplier disappears.
RFC 2644, published in 1999 as BCP 34, changed the recommended router default: routers must not forward network directed broadcasts unless specifically configured to. Cisco disabled directed broadcasts by default from IOS 12.0 — the command reference states the default as "all IP directed broadcasts are dropped". Hosts also began ignoring broadcast echo requests by default, controlled on Linux by `net.ipv4.icmp_echo_ignore_broadcasts`, which ships enabled.
With both the router forwarding and the host replies off by default, the pool of usable smurf amplifiers collapsed. The attack is now mostly of historical interest — but only because of a deliberate, internet-wide default change, not because attackers lost interest in amplification.
- `no ip directed-broadcast` on routers — the default since Cisco IOS 12.0, codified by RFC 2644 / BCP 34
- Hosts ignore broadcast echo by default (`icmp_echo_ignore_broadcasts = 1` on Linux). To confirm on a Linux host, run `sysctl net.ipv4.icmp_echo_ignore_broadcasts`; it should print `= 1`, and a 0 means the host will answer broadcast pings and could act as an amplifier
- Ingress filtering (RFC 2827 / BCP 38) stops the spoofed source packets that make reflection possible in the first place
- The same defences blunt the fraggle UDP variant, which abuses the echo and chargen services instead of ICMP
04The fraggle variant
Fraggle is smurf with UDP instead of ICMP. Rather than echo requests it sends packets to the broadcast address aimed at the UDP echo service (port 7) or chargen (port 19), both of which generate a reply. The amplification logic is identical; only the protocol changes.
It died the same way, with the same fixes: disabling directed broadcasts removes the broadcast multiplier, and disabling the obsolete echo and chargen services on hosts removes the responders. Those services are relics almost nobody runs today, which is why fraggle is even rarer than smurf.
05How today's attacks inherited the idea
Smurf is gone but its blueprint is everywhere. Modern reflection attacks swapped broadcast domains for misconfigured, internet-facing UDP services — DNS resolvers, NTP servers, memcached, SSDP, CLDAP — that each turn a small spoofed request into a large reply aimed at the victim. The strongest of them amplify far more than a broadcast segment ever did: memcached has reached five figures.
The defences are also the direct descendants of the smurf fixes. Source-address validation (BCP 38) stops the spoofing that reflection needs; closing or restricting open reflectors removes the multipliers, exactly as disabling directed broadcasts did. The full catalogue of today's reflectors, with factors and worked sums, is in amplification attacks explained and the specific case of DNS amplification.
06Smurf attack timeline
Smurf faded from a live threat to a historical footnote within a few years, and the decisive steps were defaults being changed rather than clever filters. The dates below come from the CERT advisory, the IETF documents and the vendor reference cited at the end of this page.
| When | What happened |
|---|---|
| 5 January 1998 | CERT/CC publishes advisory CA-1998-01 after reports from network and internet service providers of continuing smurf attacks |
| Cisco IOS 12.0 | Cisco changes the router default so that all IP directed broadcasts are dropped |
| August 1999 | RFC 2644 (BCP 34) requires routers to block directed broadcasts by default, updating the router requirements of RFC 1812 |
| 13 March 2000 | CERT revises CA-1998-01 to point operators to RFC 2644 / BCP 34 |
| May 2000 | RFC 2827 (BCP 38) sets out ingress filtering, which removes the spoofed sources reflection depends on |
| January 2014 | CISA alert TA14-017A lists CharGEN — fraggle's old responder — as a UDP reflector with a factor of 358.8 |
Questions
Frequently asked questions
Sources
Sources and verification
Amplification figures quoted from Craig Huegen's smurf white paper; the directed-broadcast default confirmed in Cisco's IOS command reference; icmp_echo_ignore_broadcasts confirmed under /proc/sys/net/ipv4 and the sysctl check run (procps) on Debian 12, October 2026; timeline dates from the CERT/CC 1998 advisory collection and the RFC headers.
- 01CERT Advisory CA-1998-01: Smurf IP Denial-of-Service Attacks · CERT/CC (SEI)
- 02The Latest in Denial of Service Attacks: Smurfing (C. Huegen) · pentics.net
- 03RFC 2644 / BCP 34: Changing the Default for Directed Broadcasts in Routers · IETF
- 04RFC 2827 / BCP 38: Network Ingress Filtering · IETF
- 05Cisco IOS IP Application Services Command Reference: ip directed-broadcast · Cisco
- 06Alert TA14-017A: UDP-Based Amplification Attacks (CharGEN, memcached factors) · CISA / US-CERT
Last updated