SLP amplification works by enlarging the reply before the attack starts, rather than by finding a service whose normal answer happens to be large. It is a reflection DDoS attack that abuses the Service Location Protocol, tracked as CVE-2023-29552 and disclosed in 2023, and a DDoS test looks at it from two sides: whether your edge absorbs it, and whether your own estate exposes the protocol that powers it.
How SLP amplification works
SLP (RFC 2608) lets devices on a local network advertise and discover services, and it listens on port 427 over both UDP and TCP. The reflection itself is the one every UDP vector shares: a forged source address sends the reply to the victim, which the amplification attack page covers.
What is specific to SLP is that a client can register services as well as look them up, and the vulnerable implementations accepted registrations from unauthenticated remote clients. So the reply can be grown on purpose. An ordinary SLP reply is about 12 times the size of the request. With the reply list filled, a 29-byte request draws a reply of about 65,000 bytes, a ratio of more than 2,200x.
That figure is the theoretical upper bound reported at disclosure, not a measured attack. It sits close to the largest a single UDP datagram can be, which is why it is a ceiling.
The preparation also persists. Registrations carry a lifetime, so a host that has been filled keeps returning the large reply to every spoofed request until its entries expire or are removed. The reflector becomes a prepared weapon rather than a passive one.
More than 54,000 exposed hosts were reported at disclosure, and the US CISA added the flaw to its Known Exploited Vulnerabilities catalogue in November 2023.
Why it matters under attack
The flood arrives in fragments. A reply of that size cannot cross a normal network path as one packet, so it lands as a train of IP fragments. Only the first carries the UDP header with source port 427. A filter that drops source port 427 catches that first fragment and passes the rest to the reassembly buffers, the same pressure point an IP fragmentation attack goes after.
Few reflectors fill a link. At a factor in the thousands, a small set of prepared hosts does the work that DNS or NTP reflection needs a large pool for. The source list arriving at the victim can be short and still saturate transit.
Your hosts may be the reflectors. SLP was built for local networks and has no reason to answer from the internet. Any host of yours that does is a reflector for someone else's attack, and it is also exposed to the registration step that prepares it.
Hardening, and what a DDoS test measures
The fixes are simple, and they belong to whoever runs the exposed hosts:
- Disable SLP wherever it is not needed, and always on interfaces facing untrusted networks.
- Filter UDP and TCP port 427 at the perimeter.
- Apply the patches your vendors issued for CVE-2023-29552.
On the receiving side, the test drives reflection-shaped, fragmented UDP from the test's own generators at the in-scope edge. It never borrows real SLP hosts to do it. Two readings matter: whether non-initial fragments are dropped upstream or reach the origin, and how long the flood runs before mitigation engages.
The other half is an authorised check of your own address ranges for anything that answers on port 427, repeated whenever a new range goes live. Where SLP sits among the other reflection vectors, and how a test combines volumetric floods with application-layer pressure, is mapped in understanding DDoS attack vectors.