Two stacked layers: a solid, check-marked SERVER layer (OS, middleware, CDN, WAF) and an unmaintained, dashed-outline SYSTEM layer (app, plugins, accounts, keys) marked with a warning sign, separated by a band reading 'no one is responsible'

Server Maintenance Is Not System Maintenance: A Year of Incident Response

A year ago I said incidents would surge in 2026. They did — and they aren't slowing down. After handling a large number of attacks and account takeovers, one pattern is unmistakable: the server was being maintained, but the system running on it was not. Why 'the host takes care of it' is the most expensive misconception in IT.

The prediction I wish had been wrong

A year ago I said plainly that IT incidents would surge in 2026. They did. And from where I sit, they are nowhere near finished.

As head of the systems division at Next Solutions, I have spent this past year facing a steady stream of attacks and account takeovers first-hand. Beyond our own engagements, I hold and analyze real logs from other cases as well. Across all of it, one pattern is unmistakable — and it is not the one most people expect.

In the overwhelming majority of these cases, server maintenance existed — at a bare minimum — but system maintenance did not exist at all.

The most expensive sentence in IT

Comparison articles about hosting providers have repeated some version of this line for as long as I can remember:

"The hosting company maintains the server, so users don't need technical knowledge."

I understand why it gets written. It sells. But it is deeply misleading, and I now meet the consequences of it in incident response rooms.

Here is what a hosting provider genuinely does:

  • Keeps the OS and middleware on supported versions
  • Deploys security software at the infrastructure layer
  • Provides a CDN, WAF options, and network-level protections
  • Keeps the machine running, patched, and reachable

That is real work, and good hosts do it well. But notice what is not on that list.

The layer nobody owns

What hosting providers do not do is secure the programs and systems running on top of the server. Two failure modes follow from that gap:

  • No application-layer security at all. The CMS, the plugins, the custom code, the admin accounts, the API keys, the dependency tree — nobody is applying security measures to any of it.
  • No harmony between the two layers. Even where measures exist, the system-level configuration and the server-level security are not tuned to each other. Settings that were supposed to protect something are simply wrong, or cancel each other out.

This is where the breaches come from. Data leaks and takeovers are happening in that unowned middle layer, not because the host failed, but because nobody on the customer side was maintaining the system sitting on the server.

To be explicit: this is not the hosting companies' fault. They do what they contracted to do. The failure is on the side of the people using them, who assumed that a maintained server implies a maintained system. It does not.

Backups that have never been tested

The same assumption shows up in recovery, and this is the part that turns an incident into a catastrophe.

In nearly every case I have handled, the backup plan was inadequate — and a restore rehearsal had never been performed. Not once.

An untested backup is not a backup. It is a hypothesis. You discover whether it was true on the worst day of your year, under time pressure, with your customers watching. By then it is far too late to learn that the dump was incomplete, that the credentials to restore it were themselves on the compromised machine, or that nobody knows the procedure.

"We asked our vendor. They had no idea."

There is one more structural problem, and it may be the one that needs the most daylight.

Many companies buy their hosting through a reseller — a creative agency or a consulting firm that bundles it with a website build. When an incident hits and the victim company turns to that vendor and asks "what's going on?", the conversation goes nowhere.

Not because the vendor is hiding something. Because the concept of system maintenance is missing from their model entirely. They sold a site and a hosting plan. There was never a party responsible for the ongoing security and operation of the system, and no one noticed the omission until it was being exploited.

There is no finished state

I want to be careful not to replace one comfortable illusion with another.

There is no perfect security posture. You do not install something and have the problem settle down automatically. There is no product that converts this into a solved state. Security is a practice of continuously learning and continuously confronting incidents as they come — because they will keep coming.

That is an uncomfortable answer for anyone who wanted a checkbox. It is also the only honest one.

What to actually do

If you run a system on hosted infrastructure, work through this:

  • Name the owner of the system layer. Not the server — the application, plugins, dependencies, accounts, and keys running on it. If you cannot name a party, that is your finding.
  • Separate the two contracts in your head. Hosting maintenance and system maintenance are different services. Confirm in writing which one you are actually buying.
  • Check the seams, not just the parts. Verify that your system configuration and your server-level security are consistent with each other. Most of the damage lives in that gap.
  • Rehearse a restore. On a schedule, to a clean environment, timed, with the procedure written down and the credentials stored somewhere the incident cannot reach.
  • Ask your reseller the hard question now. "Who is responsible for maintaining the system on this server?" Ask it today, not during an incident.
  • Budget for continuity, not a purchase. Treat this as an ongoing operational capability with people attached, not a one-time install.

Where we stand

At the Next Solutions System Solutions Division, we intend to hold this posture and keep expanding our capacity — maintaining stable operation of the systems we are responsible for, and developing new business alongside it.

We are hiring. If this is the kind of work you want to do seriously, we would like to talk.

日本語まとめ:「サーバー保守」はあっても「システム保守」がない

予言は当たってしまった

一年前に「インシデントは2026年に急増する」と宣言しました。その通りになり、そしてまだまだ続きます。

私はネクストソリューションズのシステム部代表として、この一年、かなりの数の攻撃や乗っ取りに直面してきました。自社案件以外の事案についてもリアルログを保有し分析しています。そこから明確に言えることが一つあります。

大半のケースで「サーバーの(最低限の)保守」はあっても、「システム保守」がまったくなかった。

「技術的な知識はいらない」という誤解

サーバーの比較記事では、古くから「サーバーはホスティング会社がメンテナンスしているから、利用者は技術的な知識がなくていい」といった表現が見受けられます。これはかなり誤解を与えていると思います。

ホスティング企業が実際にやっているのは、OS やミドルウェアのバージョン管理、セキュリティソフトの導入、CDN の提供といったインフラ層の仕事です。これ自体は確かな仕事です。しかし、ここに含まれていないものがあります。

誰も持ち主のいない層

ホスティング企業は、その上に乗るプログラム&システムのセキュリティ施策までは展開しません。結果として二つの失敗が起きます。

  • アプリケーション層の対策が皆無 … CMS、プラグイン、独自コード、管理アカウント、API キー、依存パッケージ——誰も見ていない。
  • サーバー側セキュリティとの調和が取れていない … 対策があっても設定が適切でなく、噛み合っていない。

情報漏洩や乗っ取りは、まさにこの持ち主のいない中間層で多発しています。

明確にしておきたいのは、これはホスティング会社が悪いという話ではないということです。彼らは契約した仕事をしています。問題は、それを利用している側が、サーバー上に乗るシステムの対策や保守をしていないことにあります。

テストしたことのないバックアップ

さらに、バックアップ計画が不十分で、リストアリハーサルを一度もしていないケースがほとんどです。

試していないバックアップはバックアップではなく、ただの仮説です。その仮説の真偽を知るのは、時間に追われ顧客に見られている最悪の日になります。その時点では、ダンプが不完全だったことも、復旧用の認証情報が侵害されたマシン上にあったことも、手順を知る人間がいないことも、すべて手遅れです。

「代理店に聞いても話にならない」

乗っ取り被害にあった企業が、サーバーホスティングの販売代理をしているクリエイティブ企業やコンサル企業に「どうなってる?」と聞いても、話になりません。隠しているからではなく、システム保守という概念自体がそのモデルから抜け落ちているからです。サイトとホスティングは売られたが、システムの継続的な運用とセキュリティに責任を持つ主体は最初から存在していなかった——その欠落が、悪用されるまで誰にも気づかれないのです。

完璧な状態は存在しない

もちろん、セキュリティ施策に完璧はありません。何かをインストールすれば自動で収まるわけでもありません。常に学習し、常にインシデントに対峙し続けるしかない。チェックボックスを求める人には不愉快な答えですが、誠実な答えはこれだけです。

実務チェックリスト

システム層の責任者を名指しできるか/ホスティング保守とシステム保守を契約上で区別しているか/システム設定とサーバー側セキュリティの整合(継ぎ目)を検証しているか/リストアリハーサルを定期的に実施しているか/代理店に「このサーバー上のシステムの保守は誰の責任か」を今聞いているか/一度きりの購入ではなく継続的な運用能力として予算を取っているか。

私たちの姿勢

ネクストソリューションズ システムソリューション事業部では、この姿勢を維持し、体勢を拡充しながら、システムの安定運用維持や新規事業開発に努めていきます。人材募集中です。

Sources

This article is based on first-hand incident response conducted over the past year at Next Solutions, together with analysis of real logs from additional cases. It is an expanded version of a post originally published by the author on Facebook.

Related reading