Security software on a company-issued laptop detected a North Korean operative within 25 minutes of his first login, after he manipulated session history files to hide malware installation. KnowBe4 isolated the device before any data was stolen, the company confirmed in July 2024.
The episode, reported by TIME, illustrates how modern endpoint detection moves from simple signature matching to behavioral analysis. Security agents record every process launch, network call, and file access, building a statistical baseline of normal activity for each device.
How observation becomes detection
CrowdStrike’s Bartley Richardson told TIME the sensor suite captures “process information, process trees, memory hierarchy, network calls, HTTP requests.” A process tree maps the parent-child relationships between programs; when a familiar application unexpectedly spawns a new process that reaches for sensitive data or an unknown server, the deviation stands out.
“We’ve observed so much normal behavior that we have a statistical sense of what’s normal and what’s not,” said Eerke Boiten, professor of cybersecurity at De Montfort University, speaking to TIME.

From signature to behavior
Traditional antivirus compares files against a database of known malicious code — signature-based detection. That remains useful, but modern threats rarely reuse exact code. Behavioral monitoring flags anomalies such as a never-before-seen binary executing, mass file encryption typical of ransomware, or a login from an unexpected country followed by bulk downloads at 2 a.m., according to Jake Moore, global cybersecurity advisor at ESET, speaking to TIME.
Correlating weak signals
Single anomalies generate false positives. A late-night download could be an employee catching up on work. Boiten told TIME the challenge is “lots of lots of things that look anomalous, but are false positives.” Effective systems correlate events: an unfamiliar login plus administrative commands plus an unknown binary plus an outbound transfer to a new destination raises the confidence score enough for automated isolation.
KnowBe4 timeline
The North Korean applicant passed four video interviews and background checks using a stolen U.S. identity and an AI-altered image, TIME reported. Once the laptop was issued, the user installed malware and altered session history files. The endpoint agent flagged the manipulation; staff confronted the hire, received an unconvincing explanation, then silence. KnowBe4 isolated the machine within 25 minutes and said no data was exfiltrated.

What the case reveals about insider risk
The attacker cleared human vetting — interviews, identity documents, background screening — but failed the machine’s behavioral baseline. Moore told TIME, “Hackers do leave digital footprints, and that’s what security software is constantly looking for.” The distinction is that the software watches actions, not credentials.
Automated containment isolated device in 25 minutes
How quickly can endpoint detection isolate a compromised device?
In the KnowBe4 case, automated containment occurred within 25 minutes of the first alert, according to the company’s public statement cited by TIME.
What is a process tree and why does it matter?
A process tree shows the hierarchical relationship between running programs. Anomalies — such as a word processor launching a scripting engine that then opens a network connection — betray malicious activity even when the code is unknown, Richardson told TIME.
Stu Sjouwerman, co‑founder of Sunbelt Software, started KnowBe4 in August 2010 to focus on security‑awareness training and simulated phishing. The company held a successful IPO in April 2021 and was taken private by Vista Equity Partners in February 2023; by July 2024 it had completed the acquisition of Egress, expanding its platform to cover email security and AI‑driven human‑risk management for more than 70,000 organizations worldwide.
Does behavioral monitoring replace signature-based antivirus?
No. Signature matching still catches known malware efficiently. Behavioral analysis adds a layer that catches novel or fileless attacks by spotting deviations from established baselines, Moore told TIME.