The Hall of Mirrors at the Edge of the Network

AI is accelerating attacks faster than defenders can patch. The answer may be to let adversaries in—and watch them wander.

By The Defense Desk

Somewhere inside a defense agency's network, a set of credentials sits waiting. They look real—plausible username, valid-format password, the kind of access token a mid-level administrator might have. But they lead nowhere useful. They are bait. And if an adversary picks them up, the defenders watching from the other side of a one-way mirror will know exactly who came in, how they moved, and what they were after. [1]

This is the logic of cyber deception, and according to Josh Salmanson, Vice President of Defensive Cyber Practice at Leidos, it may be one of the most underused tools in the defender's kit. 'The U.S. military has used deception in warfare forever and yet we haven't really applied it in cyber,' Salmanson told Breaking Defense. [1]

The problem is not merely tactical. It is structural—a mismatch between the speed at which threats evolve and the pace at which institutions can respond. The architecture of defense has not kept up with the architecture of attack.

The threat driving this urgency has a name Salmanson keeps returning to: the hockey stick. 'The volume and velocity and variety of attacks have changed dramatically and it gets worse every year,' he said. 'It's an absolute hockey stick right now from the exponential explosion of attacks.' [1]

At the center of this acceleration is AI. Paul Welch, Senior Vice President and Business Area Lead for Defense Agencies at Leidos, frames it simply: 'Let's start with the 800-pound gorilla: AI.' The use of AI by threat actors—both criminal and state-sponsored—is increasing risk across cyberspace. [1]

What makes the AI-enabled threat distinctly difficult is its capacity to surface problems defenders did not know they had. 'The things that we're seeing right now, especially with the emergence of Mythos-class models and some other capabilities, suggest that operational tempos for defensive cyber activities may need to increase substantially,' Salmanson said. 'What agencies and industry are seeing is these advanced models can identify vulnerabilities and exploit paths at speeds and scales that were previously difficult to achieve. In many cases, they are rapidly surfacing weaknesses that organizations did not realize existed,' he said. The result is a forced acceleration: 'defenders will likely be forced to patch things faster,' and that pressure will 'put a tremendous strain on the operational and engineering teams.' [1]

Speed, however, is not the only dimension under pressure. Trust is too. When defenders cannot be certain what is real inside their own environments, the adversary gains an asymmetric advantage that no patch can fully close.

This is where the concept of identity as perimeter becomes critical. 'Identity is the new perimeter, and there are many, many more identities than people that use the systems now,' Salmanson said. Organizations must manage system accounts, automated services, AI activities, API integrations, and cross-cloud connections—all of which expand the attack surface and blur the boundaries of what is truly inside the environment. [1]

With so many identities in play, distinguishing legitimate activity from adversarial behavior becomes significantly harder. The classical idea of a defended perimeter—a firewall, a gate, a clear inside and outside—has dissolved. In its place is something more like contested territory, where the most effective way to identify an intruder may be to introduce 'carefully designed deceptive elements into the environment—decoy credentials, false data, synthetic services or other lures that appear operationally valuable to an attacker.' [1]

The obvious objection is that deception requires an adversary to already be inside. Welch addresses this directly: the goal is not to stop every entry, but to channel it. 'You can channel and direct where an adversary goes by making a target appear attractive, interesting, and more accessible to them. But where they're actually going or potentially going is ultimately somewhere that is not useful to them.' [1] The adversary wastes time on false data while defenders observe their tools, tactics, and procedures in real time.

Salmanson extends the metaphor: 'The adversaries think they're still in the real environment, yet they're in a hall of mirrors that's totally under our control. If they attempt to get back into the real environment, we know how they got in the first time, because we can go back and track all of that forensically.' [1] Deception, done well, is not passive. It is intelligence collection by another name.

But even the most elegant deception strategy cannot compensate for infrastructure that was never designed to absorb this kind of pressure. The deeper problem is one of inheritance—systems built for a different era, now expected to hold against threats their architects never anticipated.

Salmanson is direct about this gap. 'The environments that almost all of us have to defend are based on older architectures that have been sustained for two, three generations now. The industry keeps making incremental progress, which isn't enough to face an exponential attack cadence.' [1] The solution he proposes is not incremental: organizations need to change the core hardware running enterprise software. 'When the adversaries are driving Ferraris and the defenders are still driving a Ford Fiesta from 1982, it is clear the defenders are going to have some problems.' [1]

The fragility compounds through a separate but related failure mode: technical debt. In complex environments, those patches can unintentionally disrupt other mission-critical functions, creating operational instability. [1]

The coordination problem is equally acute. [1] Seams—the gaps between adjacent warfighting forces or network domains—become operational vulnerabilities. 'Any time you have a warfighting force operating next to another warfighting force there is a potential gap in those forces' operational areas,' Welch said. [1]

Leidos frames its own positioning around this convergence of speed, resilience, and mission integrity. [2] Whether that framing meets the scale of the challenge Welch and Salmanson describe is a question the next generation of attacks will answer.

What emerges from this picture is less a technology problem than a tempo problem—a collision between the speed of institutional change and the speed of adversarial adaptation. The defenders who survive it will be the ones who designed for failure, not just for strength.

The hall of mirrors only works if defenders build it before the adversary arrives. That means resilience by design, hardware that matches the threat, and the intellectual honesty to admit that the perimeter is already gone. The question is not whether adversaries will get in. It is what they find when they do. [1]