Home › Topics › DDoS Detection & Response › Calling the ISP

When (and How) to Call the ISP During an Attack


There is one phone call that actually ends a volumetric denial-of-service attack, and most administrators have never made it. It goes to your internet provider. That is the organization whose pipe is bigger than yours and who sits upstream of the place where your circuit is being filled. Everything you do on your own equipment during a flood is, at best, damage control. The provider is the only party who can make the flood smaller before it reaches you.


That call is also, for many small teams, a mystery. Who do you dial? What do you ask for? What will they want from you, and how long will it take? This guide answers those questions from the defender's chair. It covers what an upstream provider can do that you cannot and the moment at which calling becomes the right move. It covers the information sheet that turns a vague plea into an actionable ticket, plus a script for the call itself. And it covers what to expect afterwards and the evidence to keep for the review.


One thing to say up front: almost everything that makes this call go well happens weeks before the attack. The contacts, the questions about what the provider offers, and the circuit details all belong on a sheet you prepared while it was quiet. If you are reading this during an attack, skip to the info sheet and the script. If you are reading it on a calm afternoon, the preparation section is the most valuable part.


In this guide

  • What upstream can do that you cannot
  • The four options and their tradeoffs
  • The trigger for calling
  • The info sheet they need from you
  • Contacts to find before the attack
  • The call script
  • What to expect: timelines, options, cost
  • Evidence to keep for the after-action
  • Where to go next

What upstream can do that you cannot


Your circuit has a ceiling. When more traffic arrives at the provider's side than the circuit can carry, the provider's router discards the excess. It does not care which packets it discards. By the time anything reaches your firewall, the damage is done. That is the whole reason the provider matters. Their equipment sees the traffic before the bottleneck, on links many times larger than yours. It can decide what to drop while there is still room to be selective.


Providers also see things you cannot. Their view of your circuit shows the true inbound volume, including everything that never made it to you. Their border routers see the source patterns across the whole flood, not just the sample that squeezed through. And if the attack is collateral — aimed at a neighbor who shares an upstream link with you — the provider is the only one who can tell you so.


The four options and their tradeoffs


Different providers offer different things, under different names and at different prices, but nearly everything falls into four generic categories. Understanding them lets you ask for what you want rather than waiting to be offered something.


Filtering at their edge

The provider adds a rule on the router facing your circuit, or at their network border, that drops a specific pattern. That could be a protocol you do not use, a source port associated with reflected traffic, or packets of a particular size. Or it could be traffic to a port on your address that offers nothing. This is the most surgical option and the one to ask for first when the flood has a clear signature. It costs you nothing you were using. Its limit is that it works only when the attack traffic looks different from your real traffic. A flood aimed at your web port that looks like web requests cannot be filtered this way without dropping customers too.


Rate limiting

Instead of dropping a pattern entirely, the provider caps it: allow this class of traffic up to a certain rate and discard the rest. It fits when the attack traffic is mixed in with a little legitimate traffic of the same kind. In that case, you keep a trickle of the real thing while the flood is held below the level that saturates your circuit. The tradeoff is that some legitimate traffic in that class is lost along with the attack. Picking the right limit takes a couple of tries.


Blackholing a target address

This is the option that surprises people. The provider tells its network, and often its neighbours, to discard all traffic to one specific address of yours. The flood aimed at that address disappears at the provider's border. Your circuit empties out, and every other address and service you have comes back to life. The catch is total and deliberate: the blackholed address is now unreachable by everyone, attacker and customer alike. You have sacrificed one service to save the rest.


Be honest with yourself and your management about what that means. Suppose the target is a public web server and the rest of the circuit carries phones, VPN, email, and a warehouse system. In that case, blackholing the web address is often the right call. One service down on purpose beats everything down by force. If the target is the address your entire organization lives behind, blackholing it is just the outage with a signature on it. Some providers let customers trigger this themselves, in advance, by tagging a route announcement in an agreed way. Ask whether yours does, because it turns a phone call into a thirty-second action. Also ask how they lift it, because the attacker may return the moment the address is reachable again.


Rerouting through a scrubbing service

The provider, or a partner they work with, diverts all traffic for your address through a facility built to absorb floods. They discard the attack and pass the clean remainder on to you. This is the option that keeps the target online during a large attack, and it is the one small organizations most often lack. It usually requires an existing arrangement, takes time to switch on, and may add some latency while active. If your provider offers it, ask ahead of time whether it is always-on or activated on demand. In that case, also ask how activation is triggered and what it costs.


OptionWhat it doesWhat you give upWhen it fits
Edge filterDrops a specific pattern before your circuitNothing, if the pattern is not legitimate trafficFlood has a clear signature (port, protocol, size)
Rate limitCaps a class of traffic at a set rateSome legitimate traffic in that classAttack mixed with real traffic of the same type
BlackholeDiscards all traffic to one address at the provider's borderThe target address entirely, for everyoneOne address targeted; everything else more important
Scrubbing rerouteDiverts traffic through a cleaning facility, returns the clean partSome latency; cost; setup time unless pre-arrangedTarget must stay online through a large flood

The trigger for calling


The rule is one sentence: if the circuit is saturated from outside, only upstream can help, so call. Confirm it in two minutes with two checks. Your WAN inbound graph is pinned at the circuit's rated ceiling. Also, either the provider-side graph (if you have access to one) shows more than you are receiving or your interface discard counters are climbing. Both mean traffic is being dropped before or at your edge. From that moment, every minute spent tuning the firewall instead of dialing is a minute wasted.


Three softer triggers are worth a call too. A protocol or application attack that has persisted through your local levers — connection limits, timeouts, rate limits — and is still growing. A pattern in the source view that the provider could filter cleanly, such as reflected replies from a service port you never use. Another trigger is any situation where your office internet, VPN, and phones are degraded but nothing you own is the target. That suggests you are collateral on a shared link, and the provider needs to know.


Call early. Opening a ticket costs nothing. A ticket opened during the first ten minutes with good data gets worked faster than one opened an hour later in desperation. You do not need to be certain it is an attack; you need to be able to describe what you see. If it turns out to be a backup job, you close the ticket and apologize, and the provider will not mind.


The info sheet they need from you


A provider's first-line support cannot act on "we are under attack, please help." Their network engineers can act on numbers. One thing separates a ticket that gets escalated immediately from one that bounces through three rounds of questions. It is whether you arrive with the following items already written down. Keep this template somewhere you can reach when the network is down — a printed sheet, a note on your phone.


DDoS PROVIDER INFO SHEET
Account / customer number:   ACME-0421
Circuit ID:                  CKT-7719-B (500 Mbit/s, fiber, main office)
Callback name and mobile:    Sam Rivera, +1 555 0142 (not the office line)

Target address(es):          203.0.113.10 (public web), port 443
Attack start (local time):   14:07, timezone UTC-5
Current inbound rate:        ~498 Mbit/s, ~84,000 packets/s (circuit pinned)
Current outbound rate:       ~4 Mbit/s
Source pattern:              thousands of sources; replies from UDP port 123
                             on unfamiliar servers; ~740-byte packets, constant
Family suspected:            volumetric, reflected
What still works:            nothing external; internal LAN fine
Local actions taken so far:  none on the firewall (evidence captured 14:12)
What we are asking for:      filter inbound UDP source port 123 to 203.0.113.10;
                             if insufficient, discuss blackholing .10
Evidence available:          WAN graphs, flow top-sources export, firewall log

A few notes on the fields. Give rates in the units the provider uses: bits per second for volume and packets per second for rate, both inbound and outbound. Give the start time with a timezone, because their ticket system will be in a different one. Describe the source pattern in terms they can turn into a filter: port, protocol, packet size, and whether the sources are many or few. Name the family in plain words. And state what you are asking for, even if you are open to alternatives, because a request is easier to act on than a description.


Contacts to find before the attack


The single most common failure in this whole scenario is spending the first thirty minutes of an attack looking for a phone number. The general support line that handles billing questions is not where floods get stopped. Find these before you need them, and keep them somewhere that does not depend on the network being up.


  • The emergency or network operations number — the line that reaches engineers rather than the ticket queue, and the hours it is staffed.
  • The abuse or security contact, if the provider has a separate one for attacks.
  • Your account manager, who can escalate when the front line stalls.
  • Your circuit IDs and account numbers, for every circuit, written on the same sheet.
  • The answers to five questions, asked on a calm day: What is your process when a customer reports a flood? Do you offer edge filtering, rate limiting, blackholing, and scrubbing? Can we trigger a blackhole ourselves? What does emergency mitigation cost, and is any of it included? What is your target response time for an attack report?

If your organization already runs a structured program for holding the provider to its commitments, this sheet belongs inside it. The guide on escalation with data describes that general program, including the support ladder and how to ask for escalation. This guide will not repeat it. Store the finished sheet in three places. Keep a printed copy near the firewall and a copy on a phone that does not rely on office Wi-Fi. Keep a copy in the DDoS runbook.


Tip: make a test call. Once a year, or whenever the provider changes, dial the emergency number on an ordinary afternoon. Tell them it is a contact verification. Confirm that the number reaches someone who knows what a flood is. A contact sheet that has never been dialed is a hope, not a plan.


The call script


The call has a shape, and following it keeps you from being routed into a generic queue. The goal is to leave the call with four things: a ticket number, a named engineer, an agreed action, and a time for the next contact.


StepWhat you sayWhat you want back
1. Identify"This is [name] at [organization], account [number], circuit [ID]. We are reporting an active denial-of-service attack saturating that circuit."Routing to the network or security team, not billing
2. Describe"Target is [address and port]. Started at [time, timezone]. Inbound is at [rate] bits per second and [rate] packets per second; outbound is [rate]. Sources are [pattern]."Confirmation they can see it from their side; their inbound number
3. Ask"We would like you to [filter this pattern / rate-limit this class / blackhole this address]. Can you do that, and how quickly?"Yes, no, or an alternative; a time estimate
4. Clarify"What will that block, what are the side effects, and how do we reverse it?"A clear statement of what you lose
5. Record"Please give me the ticket number and your name. What is the best number to reach this ticket directly?"Ticket, engineer, direct number
6. Schedule"When will you call back, or when should I call you? I will send the graphs and top-sources export to [address]."A next-contact time; where to send evidence
7. Verify(After the action) "We are watching our inbound graph. Can you confirm when the filter is live?"Timestamp of the action, for your timeline

If you can, stay on the line or in the ticket while the action is applied and watch your own inbound graph. The moment it falls off the ceiling is the moment you know the action worked, and the timestamp goes straight into your timeline. Ask what they see from their side afterwards, too. Their post-filter numbers tell you the true size of the attack and whether it is still going on, invisibly, upstream of you.


What to expect: timelines, options, cost


Expectations matter here, because the wait feels long and it helps to know what is normal. The first response is often a front-line agent gathering the information on your sheet and creating a ticket. That is why having the sheet ready compresses the step from twenty minutes to two. The ticket then reaches a network engineer, who will look at your circuit from their side. The engineer will confirm what they see and either apply an action directly or propose one. Depending on the provider's size, staffing, and the hour, this can be a matter of minutes or of hours. A well-prepared ticket and a specific request move faster at every stage; a vague one waits at every stage.


Actions differ in speed. A blackhole, if the provider supports it, is usually the fastest thing they can do, because it is a single routing change. An edge filter or rate limit takes a little longer, since an engineer has to write and apply it. The filter or limit may need a second pass to get the pattern right. Rerouting through scrubbing is the slowest unless it has been set up in advance, and it may not be possible at all on short notice.


Cost comes in a few generic shapes, and the time to learn which one applies to you is before the attack, not during the call. Some providers include basic mitigation with the circuit. Some charge per incident for engineer time or for the mitigation itself. Some offer protection only as a subscription, always-on or on-demand, with different tiers for different capacities. Some do not offer scrubbing at all and will suggest a third party. None of these is wrong; what is wrong is discovering the answer at three in the morning with the director on the other line. If cost approval is needed during an incident, know in advance who can give it.


Two other things to expect. The provider may ask you to make changes on your side — a route announcement, a configuration on your router. You should be ready to do so with the timeline open. And the attack may simply stop on its own, sometimes minutes after it started, sometimes because the attacker moved on. When it does, leave the mitigation in place for a while and agree with the provider on when to lift it. A lifted filter is an invitation.


Evidence to keep for the after-action


When the flood has passed, resist the urge to be done with it. The after-action review is where the preparedness decisions get made, and it needs evidence you can only gather now. Keep all of the following in one folder, labelled with the date and the ticket number.


  1. Your graphs for the whole window: WAN inbound and outbound, packet rate, discards, firewall sessions, and application response, with time axes visible.
  2. The provider's view: their inbound numbers before and after each action, and their description of the source pattern. Ask for it explicitly; it is the only measurement of the attack's true size.
  3. The ticket history, exported or copied: every message, every timestamp, every action taken and who took it.
  4. Your timeline, exactly as it was kept, including your own changes and the times they were reversed.
  5. The attack profile: target addresses and ports, protocols, source ports, packet sizes, source count, and how the attack changed over time. This becomes the filter the provider applies faster next time.
  6. Any extortion or threat messages, with full headers, untouched.
  7. The cost: what the provider charged, what the outage cost the organization in hours and lost business, and what management decided as a result.

That folder serves three audiences. It gives the preparedness discussion real numbers instead of guesses about how large a flood might be and how long the provider takes. It supports any insurance claim or law enforcement report. And it becomes the basis of a fair conversation with the provider about what worked, what did not, and whether their process needs to change. That conversation goes much better when it is built on their own ticket log and their own numbers.


Where to go next


You now know why the provider is the only party who can shrink a volumetric flood. You know the four things they can do and what each costs you. You know the moment at which calling becomes the right move, the sheet that makes the call short, and the evidence to keep afterwards. The response playbook shows where this call sits in the first hour. And preparing before the flood turns the contact sheet and the five questions into a standing part of your preparation. That preparation also includes capacity knowledge and drills. Both guides live under DDoS Detection & Response.


Put this into practice with SysPeep Monitor

The provider will ask for numbers: inbound bits and packets per second, when it started, what it looked like before. Those come from interface counters collected over SNMP and kept as history. SysPeep Monitor gathers them from one Windows machine. It graphs them on a shared time axis and alerts when inbound approaches the circuit ceiling. So the info sheet in this guide can be filled in from a screen rather than reconstructed from memory. The free trial is a quick install.

Download Free TrialLearn About SysPeep