It Feels Like the 1990s Again in Cybersecurity

It Feels Like the 1990s Again in Cybersecurity

I've honestly lost count of the zero day alerts lately. They just keep coming, one right after the other, and nobody seems ready when the next one hits. That's what got me thinking about the 90s.

Back then we had the same basic problem. Stuff was getting broken constantly and most people were scrambling. The difference, and it's a big one, is almost nobody was really online yet. A few businesses were. Most weren't, not in the way we are now. So when something went wrong the blast radius was just smaller.

Now? Everybody's online. Everybody depends on the same systems to run their business, take payments, talk to customers, all of it. One zero day doesn't stay in one place anymore.

We're Not Prepared. At All.

When I say unprepared I mean it in every way imaginable, which I know sounds dramatic until you look at how fast this is moving.

AI can iterate over exploit possibilities at a speed that's just insane. I don't think we've really wrapped our heads around that yet. We're finding zero days that've been sitting in software for decades. Decades. And it's not obscure software either. It's PHP. It's Nginx. It's the stuff underneath almost everything.

We weren't secure before. We just weren't being tested like this.

At some point we've got to adapt. I don't know if that looks like new languages or new protocols or what, exactly. But we definitely need to start thinking about what the spec for the post AI internet needs to look like, because the current one was built for a different world.

A Hard Break, Not More Patching

Here's where I probably lose some people. I think we need to start from the ground up with newer technologies and have a hard break with anything older than a date everyone agrees on.

Pick a date. Draw the line. After that, we stop pretending we can keep dragging this old foundation forward forever.

Rust is a great example of this. If we can make business owners understand that languages like Rust are going to be necessary and not legacy languages like C and C++, that would be a huge milestone. Most owners I talk to don't care what language something's written in. I get that. They care when it breaks. They care when it's in the news. By then we're already having a very different conversation.

The other piece is pentesting, and I'm not talking about a team you call in at the end to check a box.

Every software company will need to have a pentesting team. Period. That team needs to work in lockstep through the whole development process, all the way through testing and deployment. I see this role being absorbed by SREs, or SREs being absorbed by pentesting teams. I don't know which way that shakes out. I just don't think those jobs stay separate.

Where the New Baseline Starts

I'm more of an optimist than a lot of people I talk to about this, and part of that is because I think we're already starting to see the baseline move, even if it's slow.

Regardless of what your AI flavor of choice is, nearly everyone uses GitHub, and it has Copilot reviewing each PR. This is where it begins. It's not some big dramatic moment. It's just, every pull request gets another look now. That changes habits over time.

We need to train juniors differently too, because technical debt is very real and it's very expensive if it's forgotten about. I've seen that bill come due.

A good place for juniors to start, in my opinion, is on the pentesting team. That way they learn how to think differently about security in the real world. When they do move to development, everyone will have had some training in that mindset and have done a tour or two in pentesting for the company. They'll write code like someone who's actually seen how code gets broken, because they have.

But What About the Cost?

I know the pushback. I hear versions of it all the time. Rewriting is too expensive. Hiring for Rust is hard. Slowing down development for pentesting costs money.

Okay. How do you quantify a data breach?

How do you quantify customer data leakage, or a ransomware attack? How do you quantify the loss of reputation when your customers stop trusting you? You can't. Not really.

If you take the lead now then you get to be a leader and a visionary. Or you get to be babysitting old technology for the rest of your career chasing bug after bug. That's the trade.

Why I'm Still Optimistic

Hackers and malicious nation states aren't the only ones with AI at their disposal. Hackers just happen to be the ones that have cut their old way of thinking first, while corporate America is still bound to the old ways because they're scared.

I'm not an AI doomer and I don't think it's going to take everyone's jobs, nor should it. But it's here whether you like it or not.

We've got the same tools they do. We just have to actually use them to build something better instead of using them to keep the old stuff limping along a little longer.

It's time to adapt or be made irrelevant.

Kris Clark | Solutions Architect | Tech Enthusiast | DIY Builder