All systems operational 12 elite locations · Europe & Asia Pay with crypto · 0 KYC
Home / Guides / DDoS Filters and Abuse Tickets on a Crypto VPS
Guide

DDoS filters, abuse tickets, and what no invoice covers

VPSbit legal still: document beside the scales of justice

Volumetric filtering rides every VPSbit plan; application floods stay yours. Answer abuse mail with data. Core $4.80/mo annual-eff, Forge $39.20.

Short answer

Volumetric filtering rides every VPSbit plan; application floods stay yours. Answer abuse mail with data. Core $4.80/mo annual-eff, Forge $39.20.

Included filtering is wider than an upsell and narrower than a promise

The expensive assumption in hosting is that a filter line item decides whether your site survives a real attack. Here the volumetric scrubbing is not a line item at all: it rides every SKU from the smallest VPS to the biggest dedicated box, with no toggle to find and no per-gigabit fee to fear. That inclusion is real, and it is also much narrower than the words suggest, which is exactly the gap this guide exists to close before an incident closes it for you.

What the edge actually absorbs is volumetric junk: SYN storms that never finish a handshake, UDP amplification that answers your IP with someone else's bytes, ACK abuse that tries to exhaust state tables. That class of traffic dies at the network border before it ever reaches your slice, on every VPS and dedicated tier, in all 12 locations. The single step up in the catalog is the L3/L7 scrubber sitting at the top of the metal ladder on a 25 Gbps uplink; every other tier ships the same edge behavior by default. Run ss -s during an event and you can watch which layer is actually working.

Inclusion does not extend to judgment. An HTTP flood built from residential proxies or a botnet of real browsers crosses the edge as legitimate traffic, because at the packet level it is. RFC 4732 makes the distinction formal: denial of service is a resource-exhaustion problem, and once the exhausted resource is your application's CPU rather than your port, the defense moves from the provider's scrubbing routers to your web server's configuration. The edge keeps the pipe open; it cannot decide which request deserves an answer.

The typical mistake is discovering this boundary during the incident. Operators point DNS at a fresh box, skip every rate limit, and then open a ticket asking why the filter is not stopping 10,000 login attempts per minute. It is not stopping them because each one is a valid POST. Baseline the expensive endpoints first: set nginx request limits, cap concurrent sessions per IP with limit_req, and log the numbers you consider normal so the abnormal is obvious at 3 a.m. instead of a mystery at 3 p.m.

Capacity still matters, and the catalog states it plainly: 1 Gbps unmetered at entry VPS, 2.5 Gbps on Prime, 10 Gbps on Titan, and 25 Gbps of scrubbed uplink at the top of the ladder. Bigger pipes buy absorption time; they buy judgment for nobody. Set the application limits first, then let the edge do what it is built for, and the two layers will cover each other's blind spots cleanly.

VPS from $4.80* — Core on the annual term (list $6/mo), one invoice for 12 months, no auto-charges. Dedicated with IPMI from $39.20/mo. Crypto invoice, no KYC.

Launch now

Your rate limiter is the real shield once the flood looks human

Here is the counter-intuitive part that experience keeps confirming: the moment an attack is worth launching against you, it will stop looking like an attack. Garbage-packet floods are cheap and the edge eats them for free. What actually hurts is traffic shaped like login attempts, search queries, and checkout posts — the shapes your own customers use every day, generated at a rate no human body could match.

Application-layer floods target the expensive endpoints deliberately. A page that costs 200 milliseconds of CPU and four database queries is worth roughly 100 times more to the attacker than a static asset, so that is the URL the botnet requests. The filter passes every packet because the packets are valid; classification happens in your stack. Tools like fail2ban, rate-limit zones in nginx, and per-IP concurrency caps declared at limit_req_zone are the working layer here, not the provider's ticket queue and not a bigger plan.

The economics favor you when you flatten the expensive paths. Caching a heavy page turns a 200-millisecond render into a 5-millisecond cache hit, which multiplies the number of bots required by a factor most booter budgets cannot fund. Measure the real cost with curl -w '%{time_total}' against your top endpoints, cache what can be cached, and leave the truly dynamic paths to strict per-identity limits so the blast radius stays small.

The classic failure is retrofitting all of this mid-flood, when the origin is already saturated and every configuration change competes with the attack for the same CPU. Rules written under load are also written badly. Decide the limits while the site is calm: which endpoints may cost what, which actions need a session, and where the line sits between a busy user and a script. Write it down where the next operator can find it.

None of this replaces the included filtering — it sits behind it. The edge removes the floods no application can survive, and your rate limits remove the floods no filter can distinguish. A whitelisted game server or a single-protocol VPN endpoint narrows the target further, which costs nothing and outperforms most products. Think in layers, and test each one before you need it.

An abuse reply is a dataset, and the desk reads for three fields

Most operators answer abuse mail in one of two ways that both end badly: silence, or a paragraph of denial. Silence converts a routine report into escalation; denial converts it into a longer report with attachments. The abuse desk on the other side is not a judge deciding your fate — it is a clerk matching evidence to an owner, and clerks move fastest for the operators who send data instead of speeches.

Read for the three things every report carries and every credible reply returns: the IP, the timestamp in UTC, and the process that owned the socket. Pull them from your own logs with journalctl -u nginx --since '2026-10-09 06:00:00', confirm the live connection with ss -tnp, and state the result in plain sentences. If the report is right, say so and describe the fix — the mailbox closed, the plugin removed, the credential rotated. If it is wrong, show the log lines that prove it.

Timestamps deserve a sentence of their own. Abuse desks work in UTC because they correlate reports across time zones, and a reply full of local-time guesses reads as noise. Keep the server clock on UTC with timedatectl set-timezone UTC, log in UTC, and quote log lines verbatim rather than paraphrasing them. The connection metadata a facility retains lasts 24 hours at most, so your own application logs are usually the only durable record of what happened on your box.

What breaks cases is arguing. Replies that litigate jurisdiction, payment method, or the moral standing of the complainant get filed as unresolved, because none of it changes the facts the desk has already matched. The 19 coins on the invoice, the absence of KYC, and the flag over the facility are all irrelevant to the person holding a log excerpt. Answer the data with data, and the file closes itself.

Treat every abuse mail as a request for three fields and a fix, and the escalation sequence described later in this guide never advances past its first step. Operators who answer this way are rare enough that the habit itself is protection. Write your reply skeleton once, keep it in the runbook, and fill in the fields while the evidence is still fresh in the logs.

A Tor exit is a complaint magnet by design, and survivable anyway

Running a Tor exit means volunteering to receive the internet's abuse mail, and the operators who last are not the ones who hide — exits are public by definition. What keeps an exit's address routable is reachability: a published contact, a stated exit policy, and a track record of answering. The network's own relay documentation says as much, and it is worth reading in full before the first complaint ever arrives.

Set the exit policy deliberately before the relay goes live. A full exit forwards everything; a reduced policy forwards the web ports and refuses the ones that generate the worst mail. The ExitPolicy lines in torrc decide this, and blocking port 25 — SMTP — alone removes the majority of spam complaints an exit would otherwise collect. Publish the abuse contact on the relay itself so reporters find you before they find the facility, and keep that contact monitored like a pager.

The middle-relay option exists for a reason. A middle relay carries Tor traffic but never forwards to the open internet, which means no third party ever has a reason to complain about your IP — the address never appears in anyone's logs as a source. Many operators run a middle relay for a year to build history and learn the operational rhythm, then decide whether an exit's mail volume fits their runbook and their travel schedule.

The fatal pattern is launching an exit, ignoring the first weeks of mail, and discovering the address in blocklists when it stops mattering to you but still matters to every other service on the box. Complaints you never answer get answered by upstreams instead. If you cannot staff the inbox honestly, run the middle relay; the network needs those at least as much, and your address stays clean by construction.

Exits are infrastructure, and the guides assume you will operate like infrastructure: documented, reachable, and boring in the ways that matter. Do that, and the abuse mail becomes a queue you process instead of a crisis you survive. The same discipline carries directly into every other complaint category in this guide, which is why it comes before them.

DMCA is copyright mail, not a switch that powers the box off

Shared hosting has trained a generation of operators to treat a DMCA notice as an instant, silent takedown with an invoice attached. The process here is different in kind: copyright mail is filed as copyright mail, read for its specifics, and answered on the merits. A notice does not reboot your server, suspend your slice, or read your files — it starts a documented correspondence that you can actually participate in.

A notice arrives with a URL, a description of the work, and a sworn statement. Your obligations start with looking at what was actually posted: remove the file if it infringes, restore it if the claim is defective, and keep the record either way. The template at the U.S. Copyright Office lays out the required elements, and a quick grep -c 'file.zip' access.log tells you how often the reported path was really fetched while stat on the file dates the upload to the minute.

Child-safety reports are the hard boundary, and they are not copyright mail in any sense. They route to the abuse address with the URLs, they are handled as the serious reports they are, and no payment method, jurisdiction, or billing tier changes that handling. Every host has a list of prohibited categories; this one is published and short, and this category sits at the top of it in every one of the 12 locations.

The common mistake is treating the notice as spam because the box obviously hosts nothing infringing — sometimes it does, planted by an upload path you forgot you exposed, an old CMS plugin, a test directory with directory listing on. Check the URL first, the log second, your opinions third. Unanswered copyright mail joins the same escalation ladder as any other ignored report, which is an avoidable way to lose an address you paid for.

Copyright process works the way the rest of abuse handling works: evidence in, evidence out, and a fix that ends the case. The notices are routine; the operators who answer them with specifics are not. Keep the logs that let you be one of them, and the entire process stays boring — which, in abuse handling, is the highest compliment available.

The prohibited list is short, and the coin does not bend it

Nineteen coins, 12 cities, and no KYC read like a gray zone if you want them to. The prohibited list is the part that does not flex: CSAM, phishing, spam, and attack traffic are out on every plan, paid with any coin, in every location on the map. The policy text does not change between Reykjavik and Taipei, and neither does the enforcement floor underneath it.

Why absolute? Because the floor is set by the facilities and the upstream carriers, not by the reseller's mood. A datacenter that tolerates attack traffic loses its transit; one that tolerates phishing loses its mail reputation for every tenant on every range it owns. The economics are identical in a Tier IV Zurich hall and a Tier III+ Tokyo one: the prohibited categories cost the operator real money, so they cost the abuser the account. Abuse is handled under the law of the facility location, which is stated per city before you order.

For spam specifically, the operational reality is blunt: a leased box that sends unsolicited mail at volume gets its IP range flagged within days, and the range matters to every neighbor on it. Mail from a VPS needs SPF, DKIM, and DMARC configured before the first send, a low daily volume, and a list that humans actually joined. That is more work than most rented-slice spammers ever want to do, which is precisely the point of the design.

The mistake is believing the billing method is a policy layer. XMR clears the same invoice BTC does, and the acceptable-use text reads identically over both. If a project needs one of the prohibited categories, no coin, term, or city will carry it — the correct action is to not order at all. The billing method changes nothing, and it is not designed to; pretending otherwise only moves the ending further down the same ladder.

Prohibitions this short are easy to comply with and expensive to test. Everything else — the adult content that is legal where the facility sits, the privacy tooling, the research traffic — is within the policy's reach by default. The next section covers the one enforcement tool that exists for the categories above, and how rarely it fires when you answer your mail like a professional.

Null-routing sits at the end of a sequence, not the start

Ask operators about null-routing and most describe a guillotine; the actual design is the last step of a ladder with four visible rungs. A report arrives, you reply with facts, you apply the fix, and the address goes back to quiet. The blackhole route exists for the case where the AUP is being broken flagrantly and the ticket sits unanswered — a last tool, not a first response, and the distinction is visible from the outside.

What escalates a case is silence measured in days, not hours. Reports queue in batches; a case with no operator reply within a business day starts looking abandoned; a case that also shows continued traffic — the same spam pattern, the same scan targets — looks abandoned and unrepentant at the same time. At that point ip route add blackhole 203.0.113.7 is the cheapest protection the network has for everyone else, and it gets applied without ceremony or a second warning.

The restore path is equally mechanical. You fix the cause, reply with the timestamp and the change you made, and the blackhole comes off with ip route del blackhole 203.0.113.7 once the desk confirms. Operators who answer with timestamps and process names rarely meet either command on their own address. The ladder is published precisely so you can see how far away from its top rung you are at every moment.

The avoidable failure is going dark: travel without a relay for the abuse@ contact, a ticket routed to an inbox you check quarterly, a teammate who left and took the mail password along. Every one of those turns a two-reply case into an escalation. Point the abuse contact at something you actually read — a monitored alias, not a catch-all — and the rest of this section stays theoretical for you.

Hardware changes the picture in your favor. When the workload justifies metal, rent a dedicated Forge at $49/mo list — $39.20/mo annual-eff — and the same answer-with-data discipline applies, with IPMI behind it: a wedged box reimages out-of-band while the ticket is still warm. The command sequence does not care what hardware you run; your response time does. A monitored inbox is cheap insurance on any tier you choose.

A narrower target beats a thicker filter

The strongest anti-DDoS control in this guide costs nothing: fewer listening services. Every open port is a request surface no filter has to be asked about, and most boxes drowned by volume run services their owner forgot about entirely. Unmetered bandwidth makes this worse psychologically — an unmetered port is not an unwatched one, and it is not a free pass to expose everything you can install.

Inventory first: ss -lntu on a fresh box lists what is actually listening, and the honest answer is usually longer than the owner expects. Then close by default — a nftables ruleset like nft add rule inet filter input tcp dport 22 reject that admits only what the workload needs, and nothing else. A game server with a community whitelist presents one narrow face; a WireGuard endpoint answers a single UDP port; a web app behind a caching front survives hammering that would flatten a bare origin in minutes.

Absorption capacity then works for you instead of instead of you. The entry VPS carries 1 Gbps unmetered, Prime 2.5 Gbps, Titan 10 Gbps, and the top of the metal ladder scrubs 25 Gbps before traffic reaches you. Order Core at $6/mo list — $4.80/mo annual-eff — for a whitelisted game or homelab box, and the included L3/L4 handles the floods that arrive while the whitelist handles everything smarter. The virtualization type underneath decides how much of that box is really yours under load.

The failure mode is the widened exception: one port for a panel, one for an old API, one temporarily for debugging, none of them documented. Six months later the attack arrives through the one you forgot, and the post-incident review cannot even name it. Re-run ss -lntu monthly, diff it against the last output, and treat every new listening line as a decision you have to defend out loud.

Narrowing the target also makes the included filtering more effective, because less legitimate-looking traffic crosses the edge to confuse the picture. Filter, then limit, then shrink — in that order — and the expensive mitigation products stay unnecessary for the life of the project. The final section lines the whole routine up in the order to actually run it, from an empty box to a steady state.

The DDoS and abuse routine, in the order to run it

The routine starts before the first packet of attack or abuse ever arrives, on a box that still has no domain pointed at it. Lay the base ruleset in nft -f — default drop on input, the services the workload actually needs admitted by name — and confirm what is listening with ss -lntu. Set the clock to UTC with timedatectl, ship your application logs somewhere durable, and baseline the expensive endpoints: request rates, concurrency, and the render cost of your own top pages, recorded while everything is calm and cheap to measure.

Launch day adds two habits rather than new machinery. Point DNS over only after the ruleset survives a reboot, because a ruleset that exists only in your shell history is not a ruleset. Then schedule the first monthly diff of listening sockets for 30 days out, while the memory of what should be open is still fresh enough to judge each line honestly. A target this narrow makes everything downstream easier.

When traffic goes strange, read before you react. ss -s tells you whether state tables are filling; the edge dashboards tell you whether a volumetric event is being absorbed upstream. If the packets are garbage, the edge is already handling it and your job is patience. If the packets are valid — logins, searches, checkouts — the fight moves to your rate limits, and the baseline numbers from step one are what make the abnormal recognizable in minutes instead of hours.

Abuse mail follows a fixed reply within one business day: the IP, the UTC timestamp, the process that owned the socket, and the fix if the report was right. Tor exits keep their published contact and exit policy current; copyright notices get the URL checked against the logs before anything else; prohibited-category reports get routed the way the policy says, immediately, without argument. Answered this way, cases close on the first pass and the escalation ladder never leaves the ground.

Steady state is a short list on a monthly rhythm: diff the listening sockets, re-read the rate limits against current traffic, restore-test one backup, and skim journalctl for the OOM or crash lines that predict next quarter's incident. Run that loop and the filter does its job unseen, the mailbox stays a queue instead of a crisis, and the address under your name keeps its reputation intact.

Will you null-route my address?

Only when the AUP is being broken openly and the abuse ticket has gone unanswered, which makes it the last rung of a ladder you can see coming: report, your reply with facts, a fix on your side, then restoration. Operators who answer with an IP, a UTC timestamp, and a process name almost never reach the blackhole stage at all.

Does the coin I pay with change the policy?

No. All 19 coins bill through the same invoice engine and the acceptable-use text reads identically over every one of them, in all 12 locations. Policy and payment ride separate rails on purpose, so XMR, BTC, and any of the USDT networks buy exactly the same obligations and exactly the same enforcement floor.

Is DDoS filtering an upsell here?

No. Volumetric L3/L4 filtering ships enabled on every VPS and dedicated tier with no fee, no toggle, and no request form. The only step up in the catalog is the L3/L7 scrubber on the top metal tier's 25 Gbps uplink, which is a hardware class, not a filter subscription you can add to a cheaper box.

What does a useful abuse reply contain?

The IP, the UTC timestamp, the process that generated the traffic, and the fix if the report was correct. That dataset lets an abuse desk match and end the case in one pass, and it is what keeps your address out of the escalation sequence instead of somewhere deep inside it. Replies that argue jurisdiction or payment method get filed as unresolved.

Do public game servers really attract attacks?

Yes — rivals, bot ranges, and commercial booter services usually find a public address within days of launch. The included volumetric layer absorbs the flood portion, vote spam and chat abuse belong to server plugins, and a community whitelist narrows the exposed surface more than any filter you can buy. Single-protocol endpoints and monthly port diffs do the rest.

Primary sources

More ways to pay

Pay with Bitcoin · Pay with Monero · Pay with USDT (TRC-20/ERC-20/SPL) · Pay with Litecoin · Pay with Ethereum · Pay with Solana · Pay with x402 · Pay with Dogecoin · Pay with XRP · Pay with TRX · Monero on bare metal · Dedicated servers with crypto · Pick the jurisdiction first.

Ready to launch?

Build the box — VPS or bare metal — create the password, pay the invoice that follows.