Loading prices…
STKR NewsSTKR News0 of 3 free this month
Future Tech

How to tell a cyberattack from a plain old outage

When your stack fails, don't immediately blame hackers. Most digital meltdowns come from bad updates and human error, and jumping to conclusions wastes critical time.

Originally on Ars Technica →
AB

Adrian Boysel

Contributor

Nov 19, 2026

4 min read

Photo illustration / STKR News

We live in an era where everyone is on edge. When a major service goes dark, the default reaction on social media is to yell about state-sponsored hackers or a new zero-day exploit. It feels more dramatic that way. But if you have been building in this space long enough, you know the truth is usually much more boring. Most of the time, it is just a developer who skipped a test or a configuration change that spiraled out of control.

The Paranoia Tax

As a founder, the biggest mistake you can make during downtime is misdiagnosing the cause. If you assume you are under attack, you trigger a specific, expensive, and stressful set of protocols. You might shut down services that are actually healthy, lock out legitimate users, and bring in forensic teams that cost five figures an hour. If the actual problem was just a corrupted database index or a DNS typo, you have just compounded your losses for no reason.

This "paranoia tax" is real. It drains the team's morale and creates a culture of fear. We need to get better at looking at the signals before we pull the fire alarm. In the world of crypto and AI infrastructure, where everything is interconnected, a failure in one obscure API can look like a coordinated assault. It usually isn't.

The Telltale Signs of a Technical Glitch

Glitches tend to have a specific signature. They usually happen right after a change. If you just pushed a deployment, updated a dependency, or changed a firewall rule, and then things broke, that is your culprit. It is rarely a coincidence that a hacker decided to strike at the exact millisecond your CI/CD pipeline finished a run.

Hardware failures also look different than attacks. A failing switch or a dying SSD will often cause intermittent packets or slow response times rather than a clean, malicious redirection of traffic. When systems fail because of load, they degrade. When they fail because of a bug, they crash. When they are being attacked, they often continue to run but start acting strangely, like exfiltrating data or showing unauthorized access logs.

Identifying the Malicious Signature

So, when should you actually worry? A real cyberattack usually leaves a trail that a simple outage does not. Look for the following red flags:

  • Unexpected lateral movement: Internal servers communicating with each other in ways they never have before.
  • Credential abuse: Admin accounts logging in from strange IPs or at 3 AM local time without a scheduled maintenance window.
  • Data spikes: If your outbound bandwidth is hitting the ceiling but your inbound traffic is normal, someone might be stealing your data.
  • Resource exhaustion: In a DDoS attack, you will see a massive surge of garbage traffic designed to overwhelm your capacity, not just a system that stopped responding.

The difference between a crash and a hack is often found in the logs. A crash says, "I don't know what to do with this data." A hack says, "I am doing exactly what I was told to do by someone who isn't you."

Why Builders Get it Wrong

The reason many founders and engineers jump to the "attack" conclusion is that it shifts the blame. If you are hacked, you are a victim. If you pushed a bad update, you are the one who messed up. It is a psychological shield. But in the long run, transparency is the only thing that keeps your users around. If you claim a hack and it turns out to be a self-inflicted wound, you lose all credibility.

In the AI space, we are seeing a new layer of complexity. Model drift or hallucinations can sometimes look like someone has poisoned your training data. Before you start investigating your supply chain for bad actors, check your temperature settings and your prompt engineering. Complexity is the enemy of security, but it is also the primary cause of downtime.

The Founder Strategy: First, Do No Harm

When the alerts start hitting your phone, take a breath. Don't start tweeting about security breaches until you have seen the evidence. Your first move should always be to check your own change logs. Did anyone touch the infrastructure in the last hour? Did a cloud provider report a regional outage? Ninety percent of the time, the answer is in your own backyard.

Keep your incident response plan split into two paths: one for technical failures and one for security breaches. Do not cross the streams until the data forces you to.

Building a resilient company means being honest about why things break. We spend so much time worrying about the boogeyman in the shadows that we forget to check if we tripped over our own shoelaces. The most sophisticated hackers in the world cannot do as much damage as a tired engineer with root access and a lack of sleep.

The Bottom Line

The next time your dashboard turns red, stop looking for a villain. Start looking for a mistake. If you find a mistake, fix it, document it, and move on. If you find a villain, then you deal with that. But conflating the two only makes you faster at failing and slower at recovering. Real builders know that the most common threat to their uptime isn't a hacker in another country—it is the mirror.


Read the original at Ars Technica →

The Brief

Stay Updated on Cutting-Edge Tech

A six-minute morning dispatch on the markets and the technology shaping them.

Free. No spam. Unsubscribe anytime.

Write for STKR

Become a Contributor

Earn $STKR for published stories on markets, protocols, and culture.

  • Earn $STKR for every published piece
  • Editorial support from the STKR desk
  • Byline visibility across the network
  • First look at the upcoming creator program
Apply to Write

Keep reading

All stories

Comments

24 reader responses