[Report] Pass-the-Device: Porting SASE/VPN "Trust" After Endpoint Compromise - CODE BLUE 2026 #codeblue_jp #codeblue2026

[Report] Pass-the-Device: Porting SASE/VPN "Trust" After Endpoint Compromise - CODE BLUE 2026 #codeblue_jp #codeblue2026

A report on the CODE BLUE 2026 session "Pass-the-Device: Transplanting SASE/VPN 'Trust' After Endpoint Compromise."
2026.11.17

This page has been translated by machine translation. View original

Hello, this is Usuda.

Everyone, are you taking security measures? (greeting)

This time, this is a report on the following session held at CODE BLUE 2026.

Even with Zero Trust, we still need to verify whether a device is truly trustworthy. That was the session.

Pass-the-Device: Transplanting "Trust" of SASE/VPN After Endpoint Compromise

Despite massive investment in Zero Trust, in many modern SASE/VPN deployments it is still possible to transplant "trust" to another location after an endpoint has been compromised. It has been found that identity, certificates, and posture (security state) checks are, in many cases, transferable to other places rather than being strictly bound to the device.

In this talk, we introduce "Pass-the-Device," a practical technique whose core methods have been proven effective in most red team exercises involving SASE/VPN solutions. We demonstrate how an attacker can establish access independent of the attacker-controlled host by exporting or reissuing certificates, replaying sessions, abusing the enrollment flow, and reconfiguring posture state. We also classify the recurring "trust hijacking" primitives across Palo Alto, Cisco, Netskope, and Zscaler, and explain under what conditions each technique succeeds or fails.

Attendees will learn how attackers can operate beyond the original host's EDR visibility while still leaving detectable traces in IdP (Identity Provider), SASE, and network telemetry. Finally, the talk concludes with practical mitigations and detection opportunities.

Speakers: Ruslan Sayfiev

Report

  • As we introduce Zero Trust, we place trust somewhere
  • In many cases, trust ends up being tied to the endpoint device
  • Introducing a case
    • The customer's device had EDR installed
    • Due to an attacker's intrusion, EDR detected it after 5 minutes
    • It should have been possible to respond, but an hour later the domain admin had been taken over
  • Attacks were detected only once
    • However, only once in 6 years
  • There are various attacks against Windows
    • Pass-the-Hash, Pass-the-Ticket, and so on
    • This time, we named it Pass-the-Device as a similar name
  • Is it appropriate to tie trust in a device to that device?
    • We wanted to investigate the limits of the product
    • Given a managed device, can trust be transferred to an unmanaged device?
    • If the same evidence is presented, won't it appear to be the same device?
    • Of course the defensive features work, but can we really tell whether access comes from a managed device?
  • The original beginning
    • Cisco VPN, 2020
    • Certificates were obtained through SAML login, but in fact access was possible even without the certificate
  • We have spent six years researching these issues
    • We have investigated VPN/SASE, Zscaler, and others
  • How we proceed
    • Starting from the point where the endpoint is infected
    • Obtain the user's session and gain SYSTEM privileges
    • Duplicate authenticated cookies and reuse them
    • With SASE, can a new device be enrolled?
    • If so, check whether access is possible from another device
    • The points to detect and verify have changed
    • How do we look at this?
    • What is trust in the first place?
      • Credentials, certificates, whether enrollment is possible, whether sessions can be duplicated, how device posture works, whether it can be bound
    • Approach
      • Investigation, trust map, route
    • Decision tree
      • Copy the session
        • Is it possible with SYSTEM privileges?
        • Is encryption done with DPAPI?
        • Is the private key in the TPM?
          • In fact, even this does not stop it
        • Copy the session as is
        • Device IDs and similar values can also be spoofed, so we collect them
      • Enrollment
        • Common with SASE
        • Register rather than establish a session
        • Since SSO is used, can the existing SSO be used?
        • Is SSO required at every login?
        • Once registration succeeds, proceed
        • How the VPN judges the duplicated item is what matters
  • Where is trust located?
    • VPN access methods
      • ID/password or SAML?
      • data.xml files can be viewed from anywhere
      • What can be seen from inside the device
        • Often in debug mode, so most can be seen in logs
        • Examine with Decompile and Burp
      • Even with certificate authentication, where it is stored matters
        • Often can be taken from files or memory
        • Only Intune and similar can store in TPM
        • However, even with a TPM, it can be used from within the device, so it is not perfect
    • Authentication path
      • Few support device posture
      • Many have unlimited lifetimes
  • Cisco
    • Single-use sessions
    • Re-authenticate every time
  • Palo Alto
    • Cookies can be used many times, but can be shortened by configuration
    • Configuration files are saved at every login
    • Easily obtained
    • Device posture is self-reported, so it can be evaded
  • Zscaler
    • If privileges are escalated to SYSTEM, privileges can be taken
    • Client can be deployed using SSO
  • Netskope
    • Trust can be taken by obtaining just some files
  • Demo
    • Cisco
      • Connect to VPN
      • Collect information
      • Certificate is in memory, so download it
      • Log in from an unmanaged user
      • CSD host error
      • Rejected by device posture
      • Evading it prompts for a password, so analyze the hash
      • After authentication, access was possible
      • Two sessions appeared for the same user
      • Changes such as UA can be detected with IR
    • ZScaler
      • Enroll via SSO
      • Enter the email address
      • Obtain cookie
      • Apply the tool and log in again, the cookie is used and authentication succeeds
      • A second device was registered
      • SSH was blocked by device posture
      • But it can be evaded
    • Palo Alto
      • Almost the same as Cisco
    • Netskope
      • Take registry values and process them, and it works
  • What to do
    • First, check how your own setup is configured
      • Is the certificate in the TPM?
      • Can SSO be easily performed?
      • Putting only the TPM in place is not enough
        • It only makes the certificate's key non-exportable
  • Why can't it be detected?
    • Because there is no EDR
  • Another example: ZPA
    • Detected with SIEM
    • However, a persistent session remained
    • Zscaler side must also be checked
    • Need to check enrolled items and delete them
  • What to do
    • Prepare additional detection rules tailored to the product used with EDR
    • Align with where certificates are located, etc.
    • Disconnect and revoke
    • Be careful because the IR team may not have access permissions to network devices
  • Summary
    • For items tied to a device, we need to verify it is truly that device
    • What access is used and whether it can be duplicated is important
    • To judge the extent of an incident, consider whether remote access was taken
    • We want to keep SIEM or related logs
    • Session revocation and disconnection are important

Impressions

It turns out we need to take care not only of simple SSO sessions but also of device enrollment itself.

We want to be careful going forward.

Share this article