
For years, security professionals have understood that defending networks, systems, and data relied on protecting identity- or credential-based accounts. Full-time employee accounts like Active Directory, along with contractor and partner accounts, have long been prime targets for attackers seeking the “Crown Jewels” of an enterprise.
Defending identity-based accounts becomes even more complex when identities and targets are no longer human. In this multi-cloud, automation-heavy environment, AI agents, automated service accounts, service APIs, automated workflows, and software bots present a new level of risk. Agentic agents can now make decisions and trigger actions at machine speed. Without a human in the loop to monitor these accounts, the risk of a breach grows exponentially should they become compromised.
Microsoft’s 2026 Digital Defense Report points out: “As threat actors target non-human identities, they are increasingly leveraging API and trusted workflow integrations to escalate privileges, access sensitive data, and operate at scale. This reflects threat actors’ adoption of emerging technology and adaptation to organizational infrastructure models while still using the same time-tested attack methods against the identity plane.” In other words, the same old attack vectors still work, but the stakes should they succeed are much higher.
Gartner’s recently introduced security category, the Integrated Security Operations Center (ISOC), addresses the sea change in modern security defense. The ISOC model runs collection, correlation, detection, investigation, and response on a unified technology platform with a single, normalized data layer. Security analysts in the SOC can now follow a single workflow regardless of which tool or signal originates from endpoint, network, identity, or cloud telemetry, which is crucial, as more threats originate from attacks on non-human accounts.
Asking the right questions: Identity Governance vs. Behavior Monitoring
Most organizations assume they have identity governance under control, with access reviews and vulnerability scans, along with regular certifications.
But the average length of time it takes for organizations to detect and contain a breach, which is now 247 days, a latency that highlights the continued problem of false positives. And then, there are the automated, identity-based accounts operating maliciously in the shadows that never trigger an alert because the attacker uses the right credentials. The model no longer matches the environment in which enterprises operate.
Today, identity governance is not enough. The difference between identity governance and identity behavior monitoring comes down to asking the right questions. Identity governance asks, “Should this identity have access?” Identity-based behavioral monitoring, instead, asks, “Is this identity acting normally?”
And to answer that question accurately, security teams need to first understand the AI agent’s intent, all of its possible behaviors, and its impact on connected systems before an automated identity can turn legitimate access into a security event.
Monitoring and Motion
Identity security is moving away from pure access to monitoring and motion. An identity can have the right credentials and the right permissions and still be compromised. It’s not just about who has access, but what those identities are doing on a consistent basis, and whether that behavior represents a threat.
For example, with an Impossible Travel alert, two logins occur thousands of miles apart in a timeframe no human could physically travel. Was a credential stolen? A session hijacked? Is automation involved? And most importantly, what did that identity do next after the login?
Those questions are at the heart of a threat. The security analyst needs full context around these two logins.
Puppets and Strings: The PHP2Shell Case
Think about a marionette. It’s the puppet that catches your eye at first. But it’s the strings that really require your attention.
Imagine your team runs its usual vulnerability scan, which detects that an enterprise server has been exposed. A short time later, a firewall or WAF notices a suspicious request, while your endpoint security tool flags a new process. Network detection also sees an unusual connection. Your identity systems know who authenticated.
Each tool sees something real on its own. The challenge for security teams is seeing how those observations connect to tell a full story, or how the marionette strings work together.
The recently disclosed PHP2Shell vulnerabilities make this point unusually well. PHP, a backbone of the web, is the programming language powering WordPress, Joomla, Drupal, and many other commonly used applications. Core PHP engine flaws inadvertently enabled remote code execution (RCE) and Heartbleed-style memory disclosure, affecting 70% of the web. CVE-2026-109618 and CVE-2026-109619 affect PHP request handling, allowing remote code execution in vulnerable environments. These vulnerabilities include memory disclosure and unauthenticated remote code execution in PHP itself, putting every affected website at immediate risk of exposure. Attackers can exploit them remotely, sometimes fairly easily. Details escaped before the planned disclosure; patches were accelerated; exploit code became public, and researchers subsequently observed exploitation attempts.
The strings of behavior tell the real story of a potentially compromised website. For example, a vulnerability scanner reports that an Internet-facing server is running an affected version of PHP. An IDS sensor detects traffic that matches emerging exploit behavior. Network telemetry shows the server initiating a connection it normally would not make. Meanwhile, endpoint telemetry reports an unexpected process. Those may appear to be four different security events.
Identity Behavior Monitoring is an important piece of the overall threat picture in the ISOC, providing context beyond “Did the user have the right login?” The emerging ISOC model can recognize these alerts as four strings connected to the same marionette.
With the ISOC approach, the most useful question isn’t, “Did one of my tools detect it?” Instead, the analyst asks, “What is my entire ecosystem telling me happened?” When using a single, overarching system that unites all of the SOC’s many security tools and provides full context, the analyst can answer that question faster and with greater accuracy.
Defending modern enterprise environments against identity-based attacks requires security teams to evolve beyond identity governance to embrace more comprehensive behavioral monitoring across both human and non-human accounts. The ISOC model empowers security teams by uniting disjointed telemetry into clear context, so they can detect and neutralize complex threats before they cause lasting damage.
Join our LinkedIn group Information Security Community!











