Enterprise defenses are tuned to catch the attacks that make noise. This year’s data shows attackers winning by making none.

According to Picus Labs’ new Blue Report 2026, which measured more than 338 million real attack simulations across actual client production environments in the first half of 2026, defenses are having one of their strongest years yet. Average prevention effectiveness climbed from 62% to 69%, matching its 2024 peak, and logging reached a four-year high of 58%.

The good news: The recovery is real. The bad news: It’s taking place almost exclusively at the perimeter.

The report’s sharper finding is what happens after that perimeter is crossed. Inside, the picture inverts: defenses that look strong from the outside turn soft, and are the softest of all against the quiet moves, the reconnaissance and credential theft that precede every serious breach.

This is a fault line that runs through the entire report.

A vulnerable interior behind a recovering perimeter

For the first time, Picus Labs measured post-compromise prevention with autonomous penetration testing: what controls actually break the attack chain once an adversary is already operating inside the network as an authenticated user. The Post-Compromise Prevention Rate was a meager 37%.

The perimeter now blocks roughly two attacks out of three; inside, defenses stop barely one in three.

But that average hides the more useful pattern. The interior does not fail evenly. It fails along one clean line. Here too, noisy actions get caught, while quiet ones don’t.

Malicious behavior running code or jumping between machines was blocked most of the time: lateral movement through service execution, using techniques like Sharp-ServiceExec and SMBExec, was stopped around 90% of the time, and UAC-bypass privilege escalation was almost as successful at around 85%.

That’s EDR doing its job, and years of assume-breach investment showing up, as hoped, in the numbers.

Meanwhile, the quiet work runs almost unopposed.

Reconnaissance, mapping the domain and enumerating shares and sessions, was the least-prevented category of all, only being stopped a paltry 10% of the time. Defenses did a little better at detecting credentials being quietly read out of memory at around 22%, and one variant, pulling secrets straight from the registry, was stopped in less than 1% of attempts.

An attacker can map the environment, harvest sessions, and read credential material with almost no resistance, getting their proverbial ducks in a row before taking any noisier action that would trip a control.

A signature catches the famous attack, not the behavior

One result captures why. The same credential-theft tool, Mimikatz, was run at the same objective three ways, and the prevention scores could not have been further apart.

Dumping credentials the classic, heavily signatured way, straight from LSASS process memory, was blocked almost every time. Pulling them from other memory locations, or reading them from the registry, was almost never blocked at all.

Same tool, same goal, same environment.

The only variable between these three was how conspicuous the route was. The LSASS path is loud in a way tools can match: a process opens a handle to lsass.exe and reads its memory, an event vendors have instrumented for years. Reading the registry never touches lsass and looks like ordinary privileged activity, so a control built for the first event has nothing to fire on for the second.

Alarmingly, that 94% is even softer than it sounds, because it was measured against one known build of an open-source tool whose recognizability lives in how it was compiled, not in what it does.

• Rename the strings a signature keys on and the hash is new.
• Load it in memory and nothing lands on disk to scan.
• Or skip Mimikatz altogether and take the same dump with a Microsoft-signed utility already on the box.

Each path ends with credentials in hand; only the thing the signature was watching for changed. A prevention score built on signatures tells you how well you catch what you have already seen, not whether you’re actually stopping the behavior underpinning it.

The data shows stealth pays off for attackers

And that gap isn’t confined to one tool.

The Red Report 2026 found attackers deliberately shifting toward stealth, and the Blue Report shows this behavior is working for them across the board. The single least-prevented technique in the entire dataset was hiding command history, stopped just 1% of the time. The behaviors defenders miss are exactly the low-noise ones that today’s evasion-minded attackers rely on.

Malware defense is slipping for the same reason.

The IOC-Based Prevention Rate, or how often security controls block known-malicious files delivered as downloads, fell to 50% this year, from 60% last year and 71% in 2024.

Signatures alone can’t keep up: VirusTotal takes in close to two million new files a day, and repacking a payload makes the indicator stale while the behavior underneath stays the same.

Indicator-based testing still matters, it’s the fastest way to confirm the edge stops what is already known, but on its own it only ever checks the paths someone already wrote a signature for. It has to be paired with behavioral testing that asks whether the action itself is being stopped.

Organizations can see the attacks they can’t stop

So far this is a story about prevention, about what gets stopped. Detection is the fallback: when a control fails to block an action, an alert is supposed to bring a human in. This year, this critical fallback barely fired.

Logging rose to a four-year high of 58%, but the alert score stayed frozen at 14%. Fewer than one in seven simulated attacks produced an alert. Read that again. Today, teams are collecting more telemetry than ever but are converting almost none of it into action.

The gap between what gets logged and what gets alerted is now a detection-engineering problem, not a collection one.

This year’s leaders were last year’s laggards

Last year’s leaders slipped and last year’s laggards climbed, often by wide margins.

Education fell 30 points in a single year to become the least-protected industry, while the sector that had been weakest a year earlier posted the largest gain in the dataset. The report’s own line for it is blunt: strong performance is rented, not owned, and it lasts only as long as the validation behind it.

Rising averages also masked where defenses lost ground against specific adversaries.

Even as the overall score climbed, prevention fell against nine of the ten hardest-to-stop threat groups, and every one of the top ransomware families was blocked less than 38% of the time, with Play collapsing from 50% to 13%.

These are well-documented actors with published playbooks, which is kind of the point: broad control improvement does not automatically translate into coverage against the specific, evolving groups that are most likely to come for you.

What the numbers are telling us

The recovery is itself the proof. Prevention climbed seven points because organizations re-tested controls that had drifted and fixed what the tests exposed, while the sectors that regressed are the ones that stopped testing. Validation’s no longer an annual audit; it’s the difference between this year’s winners and losers, and it only holds while it runs.

That points to three moves:

• Validate exposure, not inventory. Prove which exposures are actually exploitable in your environment instead of spinning your wheels cataloging theoretical ones.
• Harden the interior against quiet actions. Test discovery, share and session enumeration, and passive credential access as rigorously as lateral movement, with detection that triggers on what an action does, not which signature it matches.
• Treat detection rules as engineering. Write them against current behavior, confirm they fire, tune out the noise, and re-validate as things change, so logs finally become alerts.

None of this is particularly exotic, and that itself is an important point. The defenses in this report already recovered the moment someone tested them.

The attacks still slipping through are the quiet ones, the reconnaissance and credential reads that never trip a signature and, this year, rarely trip an alert either.

Read the full report

The 94% and the 3% are one finding from one experiment. The Blue Report 2026 runs the same test across the whole attack surface: prevention and detection broken down by industry and region, the MITRE ATT&CK techniques defenders block least, and the threat groups and ransomware families that prevention lost ground to this year even as the overall average rose.

If the loud-versus-quiet split in this article looks familiar, the report is the fastest way to find where that gap sits in an environment like yours, and which quiet actions you should be looking for today.

Download the Blue Report 2026 to see where your defenses are likely to hold, and where the quiet gaps are most likely to be.

Note: This article was written by Sıla Özeren Hacıoğlu, Security Research Engineer at Picus Security.