It’s hard to imagine better security, but we need to

How many breaches have there been this year? It doesn’t matter exactly, but it was too many, says Jason Walsh

For a long time, computer security was, for most people, overblown waffle. It’s hard to imagine now, but once upon a time companies selling anti-virus software were accused of creating malware in order to drum up custom. Conspiracy theories and folk tales aside, this shows how deeply unserious most people were when it came to security.

The fact that there was nothing much to secure was a big part of this, I admit. I’m not all that worried about locking the garage the Ferrari I don’t have is in.

Today, of course, we all take it very seriously. I’m not saying we do it right, but we do care. We get annoyed by it and fail to do it properly, but not for a lack of will to do it right.  

 

advertisement

 

Recently, though, I’ve come to wonder: have we stepped into a new, third era?

Artificial intelligence (AI) is, obviously, what I am getting at. Large language models (LLMs) have already taken an astonishing amount of software apart, twisted it around like a Rubik’s cube, and found the, often catastrophic, engineering flaws in it, but agentic AI’s various attacks, whether accidental or criminal, appear to have really blown the doors off things.

It would be a mistake to pin the blame on all breaches on advances in AI as, no doubt, traditional approaches to cyber crime persist. Still, recent breaches at Chainalysis, Barracuda and, closer to home, the credit unions, whatever the cause there turns out to be, do rather focus the mind. 

In fact, I can’t think of a week in the past few months that hasn’t seen some breach in the news. 

That is, I suspect, not quite accurate, and it is likely that memory is blurring incidents together over stretches of time. But there have been  a lot of incidents to blur, haven’t there? “Count the stars in the sky/Measure the waters of the oceans with a teaspoon”, as composer Philip Glass suggested.

Why do I consider this a new era, though? The first clue is that I appear to have filed it as such in my head without reasoning it through. This is why, now that I do think about it, I find it so plausible: it at least appears to have happened beneath the level of conscious attention. 

Backing this up is that I can easily imagine a world where we blundered along as we were. If you don’t include agentic AI, that is. If you do include agentic AI in the imagining, though, it seems to me that such a possibility dissolves into thin air.

So if you want to say that I’m making the whole thing up in my head, I say: yes, I am, but hear me out.

Non friction

AI’s ability to do the bits no-one wants to do, which is to say the hard, technical bits, makes cyber crime, frankly, too easy. Not that I am offended on behalf of the good, old-fashioned criminals of yesteryear here or suggesting they were some sort of respected guild; it’s just that the logical endpoint of making the barrier to entry zero is, well, horrific.

I don’t really have any interest in pointing the finger. I do have an impression, though, that we’re going to have to rethink IT security and that will take a long time, so for now we need to deal with the problem at every level, from individual users to insurance, legal obligations and technology. If a rethink is needed, and possible, then these will have to continue while it’s being done.

But what would a rethink consist of? We all agree passwords need to go, but no-one agrees on what to replace them with. We all say too much data is collected and then complain when we can’t get the tea leaf reading we want from whatever dataset is masquerading as a shop today. We’re always sticking new services on products, from consumer level right down to the network level, when we all agree that, listen Dave, I think the core stuff needs a bit of attention…

Given the recently discovered insecurities in PHP, found with AI assistance after 10 years of human reviews failed, it seems to me that core computing functions, possibly down to the level of silicon, and certainly at operating system, programming language and compiler level, will need to be thoroughly examined. Authentication, too, seems an obvious area for serious tightening of standards and better integration.

On the legal and compliance side we need to see clearer chains of responsibility, and if that means a slowing of ‘innovation’ in service provision, well, hard cheese. Cuts can be made any time, but fixing things is time-sensitive. 

In fact, if one good thing is to come out of this nightmare, perhaps it will be a final nail in the coffin of the idea that we can just keep using technology to slash the bill for wages, capital equipment, transport and all of the other annoying and boring stuff that businesses seem bent on trying to get rid of.

RELATED ARTICLES
Sign up for the
Technology Minute

Listen to Tech Radio

- Advertisment -

Most Popular