Recent major outages are due not to cybercriminals but network administrators’ failure to maintain their systems properly, authorities have argued, as the outage of a top-secret US National Security Agency (NSA) network extends a growing streak of unusual resilience failures.

The sustained outage of NSANet, an internal network that employees at the US agency use to communicate top-secret information, saw the NSA shut down its Maryland headquarters, advise staff they could return to work, then block access again as specialists worked to resolve the issue.

Amidst the ongoing Iran conflict NSA attributed the “temporary degradation” to “scheduled routine maintenance” but said its operational capabilities were intact, citing its “multiple layers of redundancy to maintain operations during planned maintenance or unexpected events.”

Threats from Iran “have substantially increased,” director of national intelligence Jay Clayton said during Congressional hearings on Wednesday, conceding that the country has “a network capability that is very sophisticated, global, and one that we need to combat seriously.”

With the outage coming days after a nationwide shutdown of the US air traffic control system, a global Microsoft outage, compromises of banking giants JPMorgan, Bank of America and others; and a JavaScript exploit that took down 178 major websites, some find the coincidences too much.

And while committee member Rep. Michael McCaul noted that just 5 per cent of threats against the US are reported publicly, Clayton said the NSA outage was due to “routine upgrades” and “we have no indication” that “external threat actors or state sponsored actors” were involved.

Pressed by committee member Rep. Dan Goldman to verify that “a series of dramatic and systematic failures in the government and private sector… is purely coincidental”, Clayton reiterated that “we have no indication that these events are related”.

Systematic coverups or just technical debt?

Yet system upgrades are hardly a new task, and the insistence of a wartime US government that cybercriminals had no role in multiple contemporaneous failures belies reports from Europe, where major cyberattacks have been openly reported and attributed to malicious foreign actors.

Recent weeks have seen cybercriminals compromise targets including Norwegian energy company Equinor, Polish e-government platform ePUAP, German drone manufacturer Quantum Systems, Ukrainian and Russian emergency alert services, Lithuania’s customs agency, and French refrigerators.

The past fortnight has also seen three signal-related train collisions in Spain, Prague and Egypt, although these have not yet been attributed to cybercriminals.

Clayton, however, echoed calls by FAA administrator Sean Duffy that the sudden spate of upgrade failures is due to a lack of funding for necessary system upgrades – suggesting that decades of resilience planning have been defeated in days by increasingly complex systems.

“The speed of compute, and the level of capability of applications that use that compute, is increasing substantially,” Clayton said, “which means legacy systems that were old 6 months ago, and a year ago, are a lot older now than they were then.”

We've been told we can expect more outages. Photo: Shutterstock

Former Cybersecurity and Infrastructure Security Agency (CISA) head Chris Krebs is willing to believe in coincidences, telling MSNOW that “I am not terribly worried but I am slightly unsettled… AI based cyberattacks are happening, but this ain’t that.”

“If it’s the Chinese [hacking NSANet] they’re either going to burn the whole damn thing down or they’re going to sit there quietly and you’ll never know they’re there,” he said, noting reports suggest “this is in that muddy middle area where it very much looks like some patch or update did not go as designed.”

Have we just forgotten how to update systems?

Managing system updates is part and parcel with running IT infrastructure, with patching a core part of the ASD’s well-established Essential Eight best cybersecurity practices and fundamental to the newly reworked ASD Essentials guidelines.

And while investigations into incidents such as the NSA outage often turn up telltale signs of compromise days or weeks after the event, Illumio public sector chief technology Gary Barlet told Information Age that patching has become a very real stumbling block in the age of AI.

Where CIOs previously went to great lengths to ensure patches didn’t interrupt service – updates were tested extensively before being deployed, scheduled for times of low usage and customers informed before deployment – Barlet said AI’s speed has made such discipline untenable.

With AI now autonomously finding holes and hacking systems – researchers this week used AI to find a decades-old flaw in the ubiquitous Java language that has been dubbed Java4Shell – “AI is making us perceive a threat and the reaction is to patch as fast as they can,” Barlet explained.

“I’m just not sure that their [updating] procedures have been refined enough to patch at the pace they’re trying to patch,” he said, and they’re unlikely to be testing patches enough “because they’re under pressure to get patches out as fast as possible.”

“People used to think they were safe because they were in an obscure industry or something that AI doesn’t care about,” Barlet said, “but now AI can be working at very fast speeds 24x7 and there’s nobody that’s not a frontline target in cyber anymore because of AI.”

Krebs agrees – noting that the unprecedented interconnectedness of today’s systems means “we are patching at a rate never before seen in history… we’re probably in the early innings of what this AI-enabled discovery wave looks like, and we’ve got to be prepared for more outages.”