SSD Nodes Learn Hosting plans →
How to do am Matt ConnorBy Matt Connor · Updated 2026-10-14

How to Stop Subscription Bombing for Signup Forms

Subscription bombing fit bury person inbox with thousands emails. Confirmed opt-in and rate limits stop your server from sending signup messages to attackers’ targets.

Wetin be subscription bombing?

Subscription bombing na attack wey use your signup form bury another person inbox. Attacker go take one victim email address submit am to hundreds or thousands unprotected forms within short time. Each website go send welcome message or confirmation message go that address. All the messages together go hide the mail wey victim really need read.

The target na person wey own that inbox. As subscription confirmations dey fill the inbox, attacker dey spend money from that person card or reset password for one of their accounts. Fraud alert from bank still go arrive. E go land underneath two thousand other messages wey enter within the same hour, so nobody go see am on time.

Your server na the tool wey attack dey use. Nothing for your box spoil. No account wey belong to you get compromised. Somebody type an address into public form, and your software do wetin dem write am to do: e send mail go that address. Na this make the attack hard to notice. No intrusion dey your logs because no intrusion happen.

Wetin the attack dey look like for your side

E dey come in one of two forms.

The loud form na burst. Several hundred POST requests go hit one form within few minutes, from plenty different source IP addresses, carrying addresses for domains wey you never send to before. This one easy to see once you check.

The quiet form na the one people dey miss. The attacker get list of thousands vulnerable forms, so your form only need contribute one or two submissions every hour. Jye Cusch describe an attack with exactly this form for one site wey e dey run: no traffic spike, just steady signups dey arrive for hours wey no match the audience. One form look harmless because one form almost no dey do anything. The damage na the total across every form for the attacker's list.

Both forms get the same signature afterward: nothing happen next. The addresses never confirm. Dem never open any message and never click any link. For confirmed opt-in list, dem remain for status unconfirmed forever, and that pile na the clearest evidence wey you go get.

Start by counting submissions per minute for your access log.

sudo awk '/POST \/subscription\/form/ {print substr($4, 2, 17)}' \
  /var/log/nginx/access.log | uniq -c | sort -rn | head

$4 for the default combined log format na the timestamp inside brackets, so this command go print count for each minute, highest one first. If form wey normally get four signups per day suddenly show sixty for one minute, that no be normal day.

Confirmed opt-in: defence wey get biggest effect

Confirmed opt-in, wey dem usually dey call double opt-in, mean say address no be subscriber until person click link for message wey dem send go that address. Turn am on, and one address wey person submit go produce exactly one message, ever. Address no go join list, so e no go receive campaign or welcome sequence.

For listmonk, self-hosted newsletter server this na setting for each list: list fit use single opt-in or double opt-in. Documentation explain the difference clearly. For double opt-in list, subscribers "explicitly accept the subscription by clicking on the confirmation e-mail they receive. Until then, they do not receive campaign messages." Subscriber dey for unconfirmed first, e move go confirmed after the click, and na only confirmed subscribers for opt-in list dey receive campaign mail.

Make you talk true about wetin this one dey solve. Confirmed opt-in no make your contribution become zero. E limit am to one message for each address. Victim still go receive that message, and one message from each of one thousand sites fit complete the attack. Wetin confirmed opt-in remove na everything wey happen after that: your list go remain clean, and you no go ever send second message to person wey no ask for the first one.

Two more settings matter, and people fit easily forget both. First, limit how many times dem fit resend the confirmation. If person fit submit the same address again and receive another confirmation email every time, attacker no need one thousand forms, because your form alone go send one thousand messages. Address wey already dey for unconfirmed on that list suppose receive nothing else for at least one day. Second, delete unconfirmed rows based on schedule. Address wey never confirm after thirty days no be pending subscriber. Keeping am only create chance say something go mail am by mistake later.

Limit di rate for signup form for inside reverse proxy

Put the limit for front of the application instead of inside am. Request wey dem block for proxy no dey ever open database connection and e no dey ever start SMTP (simple mail transfer protocol) conversation. Limit wey dey inside application dey run after request don already cost you worker process and query, and for plenty stacks, dem dey queue message before any abuse check run. Proxy limit still dey survive application upgrade, because e no dey live inside code wey you dey replace.

The example wey dey below na nginx. The idea dey work for any reverse proxy wey you dey run for front of your app, even though the directive names fit differ.

Put this one inside the http block, for file like /etc/nginx/conf.d/signup-limit.conf:

map $request_method $signup_key {
    POST    $binary_remote_addr;
    default "";
}

limit_req_zone $signup_key zone=signup:10m rate=2r/m;
limit_req_status 429;
limit_req_log_level warn;

The map dey do real work. nginx no dey count request wey e key na empty string, so na only POST requests dey enter the zone. Reader wey dey load signup page plenty times no dey spend anything. Without that map, person wey refresh page two times go burn im own budget before e even submit anything.

$binary_remote_addr na the client address for packed form. For 64-bit system wey any modern VPS dey run, nginx dey store 128 bytes of state per tracked address, so 10 megabyte zone fit hold roughly 80,000 of dem. rate=2r/m allow one submission every thirty seconds. limit_req_status 429 dey return HTTP 429 Too Many Requests instead of nginx default 503, wey be the honest code and the one wey client library dey expect.

Then inside the server block for your site:

location = /subscription/form {
    limit_req zone=signup burst=3 nodelay;
    proxy_pass http://127.0.0.1:9000;
    proxy_set_header Host $host;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}

burst=3 nodelay dey allow person wey double-click the button pass. nginx dey pass the first request plus three more at once, and e dey reject the fifth one straight away instead of make e queue am.

sudo nginx -t && sudo systemctl reload nginx

nginx -t suppose print configuration file /etc/nginx/nginx.conf test is successful. Now submit the form five times quick-quick and watch the error log:

sudo tail -f /var/log/nginx/error.log

Blocked request dey write one line, and this na the string wey you dey look for:

2026/08/13 09:14:22 [warn] 812#812: *4412 limiting requests, excess: 3.400 by zone "signup", client: 203.0.113.10, server: news.example.com, request: "POST /subscription/form HTTP/1.1", host: "news.example.com"

If no line show at all, e mean say the limit no dey apply. The common cause na say limit_req dey sit inside location block wey request no dey ever reach, so check with curl -si -X POST https://news.example.com/subscription/form five times for row and confirm say the fifth one return 429.

Two traps dey wey you suppose know before you rely on per-IP limit.

Behind CDN or another proxy, $binary_remote_addr na that proxy. Every visitor dey land for one bucket, so the first few submissions of each minute go lock out everybody else. Fix am with the real IP module: set_real_ip_from for each of your CDN published ranges (Cloudflare dey list dia own for cloudflare.com/ips) and real_ip_header CF-Connecting-IP. Confirm the fix by reading $remote_addr for your access log and checking say e be visitor address instead of your CDN own.

IPv6 dey make per-address limit weak. $binary_remote_addr dey hold the full /128, and residential IPv6 allocation usually be /64 or larger. That one plenty pass the addresses wey attacker fit work through, each one get im own clean budget. Add second zone as ceiling for the endpoint itself, keyed on constant, so the form get total rate no matter how many source addresses dey play:

map $request_method $signup_total_key {
    POST    "signup";
    default "";
}

limit_req_zone $signup_total_key zone=signup_total:1m rate=30r/m;

Add limit_req zone=signup_total burst=10 nodelay; to the same location. Set the rate above your busiest real hour with room to spare. This na blunt control: during attack e dey turn away real signups too. That na the correct trade, because the alternative na your server wey go dey send the mail.

Why per-address limit no fit dey for proxy

The email address dey inside the POST body, and nginx no dey parse request bodies. Every variable wey limit_req_zone fit use as key dey come from request line, headers, or connection. So rule like “this address fit receive only one confirmation per day” gatz dey inside the first component wey read the body, wey be your application.

No try bypass this by moving the address go query string so $arg_email go become available. This one go write every subscriber address for your access log as cleartext, and for any log shipper wey dey downstream from am. You go exchange rate limit for privacy problem.

One real exception dey. nginx JavaScript module, njs, fit read request body and set variable from am. This one allow you build per-address key for proxy. E be real option, but e also add new code to your request path. For most sites, per-address cap suppose dey near the database wey already know whether this address get pending confirmation, while proxy handle per-IP and per-endpoint limits wey e good at.

No repeat the text person submit inside the message wey you send

Keep every string wey attacker supply out of the message wey you send. Two separate reasons dey, and attackers don use both for real attacks.

If your confirmation email greet reader with name wey come from the form, attacker fit write their message inside the name field. Your server go then deliver that text to the victim from your domain, signed with your DKIM (DomainKeys Identified Mail) key. Your site don become delivery service for another person abuse, and the receiving provider go see your domain for the message.

The second reason worse. If you join any submitted field into mail header by hand, newline character inside that field fit add headers wey attacker choose, including Bcc. Modern mail libraries reject newlines for header values. Code wey pipe text go sendmail from shell script often no dey do that.

Safe confirmation message get your site name and one link, with one sentence of explanation. The address itself suppose appear only where mail transfer agent need am, inside the To header. Test am like this: submit the form with a name field wey hold newline and obvious link. Then read the raw message wey you receive with less, and check say none of dem remain.

As you dey do this, make the success page show the same thing for every address. Page wey say "you don already subscribe" for one address and "check your inbox" for another one fit turn your form into membership checker for anybody wey get list of addresses to test.

Which bot check you suppose use?

Choose for accessibility as hard as you choose for effectiveness. Image-selection captcha, blind reader no fit solve am, and the audio fallback dey hard for people wey get ordinary hearing. Check wey dey cost legitimate person dia signup na defence wey still be cost. Four options, for the order wey you suppose try dem.

Proof of work for browser. The browser dey compute hash wey server fit verify cheaply, and nothing dey for person to solve. listmonk dey offer this one under Settings, then Security, using ALTCHA, wey no need any third party service. listmonk add ALTCHA for version 5.1.0 and mark im older hCaptcha option as deprecated for the same release. The cost dey fall on whoever submit the most, wey be the attacker.

Managed non-interactive check. Cloudflare Turnstile no dey show most visitors anything at all and e dey only challenge when im signals look bad. E dey effective, and e dey put third party for your signup path.

Honeypot field. Text input wey person no dey ever see and naive bot dey fill. Give am name wey your form no dey otherwise use, and set autocomplete="off", tabindex="-1" and aria-hidden="true" so password manager no go fill am and screen reader no go announce am. Field wey dem name email2 or address dey get autofilled by browser, and then you go reject real people.

<div style="position:absolute; left:-9999px;" aria-hidden="true">
  <label for="hp_ref">Leave this field empty</label>
  <input type="text" id="hp_ref" name="hp_ref" autocomplete="off" tabindex="-1">
</div>

Time-to-submit check. Put signed timestamp for hidden field when page render, and reject submission wey arrive less than two seconds later. Person no fit read form and type address that fast. Sign the timestamp, or the bot go just send old one.

One thing to verify whichever one you pick: dem must consume the token once. If script fit solve the check once and replay that token against thousand addresses, the check prove say browser run once and nothing more.

How you fit find out before abuse report land?

You want your own graphs to tell you, no be hosting provider abuse desk. Monitor two things.

Count submissions per source address across the log:

sudo awk '/POST \/subscription\/form/ {print $1}' /var/log/nginx/access.log \
  | sort | uniq -c | sort -rn | head -20

Then make fail2ban read the same limiting requests lines wey nginx already dey write, and ban repeat offenders. fail2ban get filter for exactly this. Create /etc/fail2ban/jail.d/nginx-limit-req.local:

[nginx-limit-req]
enabled  = true
filter   = nginx-limit-req
port     = http,https
logpath  = /var/log/nginx/error.log
findtime = 600
maxretry = 10
bantime  = 3600
sudo systemctl reload fail2ban
sudo fail2ban-client status nginx-limit-req

The status output go list the jail filter and the failed and banned counts wey e get now. Currently banned: 0 for a quiet day dey correct. If the jail no show at all, fail2ban no load the file, and sudo fail2ban-client -d | grep nginx-limit-req go dump the configuration wey e actually parse. The shipped filter dey match every limit_req zone. Narrow am to your signup zone by setting ngx_limit_req_zones = signup inside a [Definition] section of /etc/fail2ban/filter.d/nginx-limit-req.local. The jail file layout and ban commands dey explained more deeply for the fail2ban guide for Ubuntu 24.04.

The second signal na ratio, and e no need new software: submissions divided by confirmations. For healthy list, most people wey submit address go click the link, usually well over half of dem. When that ratio drop sharply while submissions dey rise, dem dey use you. Compare the count of unconfirmed subscribers wey dem create for the last hour with the count of confirmed ones, for any schedule wey you already dey use run reports.

Wetin e go cost you: sender reputation and blocklists

Na this part dey turn small wahala into bill.

Dem dey harvest the address lists wey dem use for bombing. Harvested lists dey contain spamtraps: addresses wey nobody ever sign up for anything with anywhere. Dem publish am only to catch senders wey dey send mail without permission. Your confirmation message reach one of dem. Some blocklist operators no need anything pass that.

Recipients wey no ask for your message no dey click unsubscribe. Dem dey click "report spam". Google's bulk sender rules, wey start to apply since February 2024, tell senders wey dey send 5,000 or more messages every day to Gmail make dem keep reported-spam rate for Postmaster Tools below 0.3%. Smaller sender no dey measured against that number, but the same complaint signal still enter the filtering decisions wey send your mail go spam folder. The fake addresses for the run go bounce hard too, and rising hard-bounce rate na separate reputation signal for every major provider.

If you run your own mail server for VPS with mailcow, the listing go land for your IP address and your domain. To delist with operator like Spamhaus, you go fill form and wait. While you dey wait, your invoices and password resets no go deliver too. If you send through shared provider instead, expect dem to suspend your account first and read your explanation later, because your traffic na risk to every other sender for that IP.

Compared with that, the work small. Turn on confirmed opt-in today, because na one setting for each list. Add the proxy rate limit next, because na one file and one reload. The bot check and alerting fit follow this week.

FAQ

Double opt-in fit stop subscription bombing?

E fit stop your list from getting polluted, and e limit your own contribution to one message for each address wey person submit. Na the biggest single improvement wey dey available to you. But e no go stop the victim inbox from filling up, because the attack na the total of one message from each of one thousand sites. Combine am with per-IP rate limit for your proxy and limit on confirmation resends, so submitting the same address twice no go produce another message.

How I fit know subscription bombing run from one good day of real signups?

Check wetin happen after submission. Real signups dey confirm, and dem usually confirm within hours. Bombing run dey leave plenty addresses wey never confirm, never open, and never click. The submissions too dey form strange pattern: many source addresses wey you never see before, recipient domains wey you no normally send to, and arrival times wey spread evenly across the whole day instead of following when your audience dey awake.

I suppose delete the addresses wey people submit?

Yes. Delete unconfirmed records wey don pass about thirty days, and do am on schedule instead of by hand. Never send anything else to those addresses, including apology or a "was this you?" message, because that one na another message wey the person no ask for, after dem don already receive plenty of them. If any of those addresses na spamtraps, follow-up na the confirmation wey the blocklist operator dey wait for.

Rate limiting go reject real subscribers?

Per-IP limit of one submission every thirty seconds, with burst of three, no go affect person wey fill form once. E go show when many real people share one address, like office wey dey behind one NAT (network address translation) gateway, or when your proxy dey see your CDN address instead of the visitor address. Read $remote_addr for your access log before you make the limit tighter, and keep the endpoint ceiling above your busiest real hour.

My sending IP dey on blocklist after one run. Wetin I suppose do first?

Stop sending from am before you request anything. Pause the campaign queue, fix the form, and delete the unconfirmed addresses, because delisting followed by more of the same traffic go make dem relist you faster than the first time. Then find the list wey carry your IP, because most operators get lookup page wey you fit use with your IP address, and follow their removal process. Expect the wait to take days, and use that time confirm say your SPF (sender policy framework) record and DKIM signing still dey pass.

Dis article help you?

#email#double-opt-in#rate-limiting#abuse#deliverability