Containment on microsegment.io

The AI Agent Was New. The Trust Failures Were Not.

Wed, 29 Jul 2026 00:00:00 +0000

The headline is an autonomous AI agent escaping an evaluation sandbox and compromising Hugging Face production.

The security lesson is less exotic.

One compromised workload could read credentials. A pod could reach cloud metadata. An overly privileged Kubernetes identity could create privileged pods. A stolen mesh-VPN key could enroll attacker-controlled devices. One shared service-connector credential provided administrative access across multiple clusters.

The agent was new. The trust failures were not.

Backup Infrastructure Needs Its Own Containment Boundary

Tue, 28 Jul 2026 00:00:00 +0000

Most backup strategies start with copies, retention, and restore speed.

Attackers start with reachability.

Can a compromised production administrator reach the backup console? Does the backup platform trust the production identity provider? Can ordinary server networks connect to repositories or storage management interfaces? Can a shared virtualization manager delete both workloads and their recovery points?

If the answer is yes, the backup environment is not outside the blast radius. It is simply another management plane inside it.

AI Data Pipelines Need Containment

Tue, 21 Jul 2026 00:00:00 +0000

Security teams are used to treating uploaded executables as hostile. Uploaded datasets often receive less suspicion.

That distinction no longer holds.

On July 16, 2026, Hugging Face disclosed an intrusion that started in its dataset-processing pipeline. A malicious dataset abused two code-execution paths: a remote-code dataset loader and template injection in a dataset configuration. The attacker reached a processing worker, escalated to node-level access, harvested cloud and cluster credentials, and moved laterally into several internal clusters.

Your Identity Provider Needs a Containment Boundary

Tue, 14 Jul 2026 00:00:00 +0000

Identity is often described as the new perimeter. That framing misses the harder architectural question: what protects the system that creates the identity assertion?

New Mandiant research makes the problem concrete. During a red team engagement, Mandiant recovered an active Active Directory Federation Services token-signing key from the host’s machine-scoped cryptographic store. With that key, the team forged a SAML assertion for a Global Administrator identity that Microsoft Entra ID accepted.

The ATT&CK Segmentation Layer Is Now Updated to v19.1

Fri, 10 Jul 2026 00:00:00 +0000

The microsegment.io ATT&CK Navigator layer for Network Segmentation is now refreshed to MITRE ATT&CK v19.1.

This is not a dramatic update in the sense of new colors, new scoring, or a long list of new techniques. The useful part is more boring, and probably more important: the mapping stayed stable. MITRE’s current Enterprise ATT&CK data still maps M1030 Network Segmentation to 44 technique and tactic entries across 37 unique technique IDs.

That stability is worth paying attention to. It means the original point from the 2019 ATT&CK segmentation article still holds in 2026: segmentation keeps appearing across the attack lifecycle because attackers need reachability. They need paths to remote services, identity infrastructure, management systems, deployment tooling, configuration stores, command-and-control channels, and exfiltration routes. If those paths are broad, implicit, or poorly governed, the attacker gets options. If those paths are narrow, explicit, and enforced close to the systems that matter, the attacker has fewer moves available.

What Good L3/L4 Policy Can Actually Do

Thu, 09 Jul 2026 00:00:00 +0000

In 2019 I published a short article about MITRE ATT&CK and segmentation.

The point was simple: when ATT&CK maps adversary techniques to mitigations, network segmentation shows up in more places than many people expect. It is not only a perimeter control. It appears across the attack lifecycle because attackers need reach.

That old point still holds.

But it deserves a 2026 reassessment.

The original post was written in a world where many segmentation conversations still meant VLANs, firewall zones, and large network boundaries. Today the conversation is broader. We have cloud security groups, Kubernetes network policies, host firewalls, software-defined segmentation, identity-aware access paths, workload labels, service maps, and microsegmentation platforms.

Voice Infrastructure Needs Containment, Too

Tue, 07 Jul 2026 00:00:00 +0000

Voice infrastructure is easy to underestimate.

It sits in the background. It routes calls. It connects desk phones, softphones, contact centers, voicemail, emergency calling, directories, gateways, and collaboration workflows. It is operational plumbing, so it often gets treated as a specialist island.

That is the wrong mental model.

Unified communications platforms are not just voice systems. They are trust infrastructure.

The latest Cisco Unified Communications Manager issue is a useful reminder. CVE-2026-20230 affects Cisco Unified CM and Unified CM SME when the WebDialer service is enabled. Cisco published the advisory on June 3, 2026 and updated it on July 1, 2026 after becoming aware of active exploitation. NVD describes the issue as server-side request forgery through crafted HTTP requests that can allow file writes to the underlying operating system, which can then be used to elevate to root. CISA added the vulnerability to the Known Exploited Vulnerabilities catalog on June 25, 2026 with a June 28 due date for covered federal agencies.

Mythos, Microsegmentation, and the Collapse of the Exploit Window

Fri, 03 Jul 2026 00:00:00 +0000

The Interesting Part Is Not the Hype

Anthropic’s Project Glasswing is easy to turn into a dramatic headline.

Claude Mythos Preview finds vulnerabilities. It helps develop exploit paths. It can be pointed at large, important software systems. In Anthropic’s first update, partners reported more than ten thousand high- or critical-severity findings in about a month.

That sounds like the future of vulnerability research arriving all at once.

Maybe it is.

Management Consoles: The Keys to the Kingdom

Wed, 08 Apr 2026 00:00:00 +0000

The Wrong Thing Is Still Trusted

Defenders keep hardening endpoints, tuning detections, and buying more visibility.

Meanwhile, attackers keep going after the systems that already have permission to touch everything.

That is the real problem with management consoles.

When a laptop gets compromised, you have an incident.

When a management console gets compromised, you may have a change-control problem, an identity problem, a visibility problem, and a lateral movement problem all at once.