
- DBSC ties each session to its device.
- It reduces off-device cookie abuse, but it does not remove malware or replace MFA.
- Start with a measured pilot and keep a safe fallback for unsupported devices.
Passwords are not the only path in. Malware can copy a signed-in browser cookie and send it to another machine. The attacker may then enter without typing the password or completing the original sign-in check.
Device-Bound Session Credentials, or DBSC, changes that model. The browser creates a private key that should stay on the device. The server can ask for proof from that key before it refreshes short-lived session cookies.
Copy the cookie, and the proof is missing.
Think of it as a door pass tied to one safe key. A photo of the pass is not enough at the next check. The guard asks for the key too. That is the core idea, even though the web process uses code and secure hardware.
Teams can test this in small steps. Keep the first trial small. Start with staff, learn from each failed check, and fix the path before more users join.
Google moved DBSC into public availability for Windows users in Chrome 146 during 2026 and said macOS support would follow. The work is also being developed as an open web standard. That makes this a useful planning point for security and web teams, not a reason to rush a full rollout overnight.
Why a password reset may not end a stolen session
Many web sessions use bearer cookies. The browser presents the cookie, and the server treats possession as proof. If malware steals it, the copy may work elsewhere until the server expires or revokes it.
MFA and passkeys strengthen the sign-in step. They do not automatically prove that every later request comes from the same device. That gap is why session theft remains useful to attackers.
Our passkey readiness checklist helps with authentication. DBSC addresses a different part of the chain: the life of the session after sign-in.
How DBSC works in plain language
At registration, the browser creates a key pair for the session. The private key is protected by device hardware when supported, such as a Trusted Platform Module on Windows or the Secure Enclave on macOS. The server receives the public part.
The server still uses normal cookies for web requests. Those cookies can be short-lived. When they need a refresh, the browser proves it still holds the private key. A copied cookie on another machine cannot create that proof.
The Google Security Blog explains that each session uses a separate key. The design aims to avoid giving sites a stable device ID or a new way to track users across sites.
That privacy goal matters. Security should not quietly become fingerprinting.
What the control can and cannot stop
| Scenario | DBSC effect | Other control needed |
|---|---|---|
| Cookie copied to another device | Strongly limits reuse after refresh | Short cookie life and revocation |
| Malware acting on the same device | Limited protection | Endpoint detection and response |
| Stolen password | Not the main purpose | Passkey or MFA |
| Risky browser extension | Does not remove access | Permission review and policy |
| Phishing before sign-in | Does not block every flow | User checks and phishing-resistant auth |
Device binding raises the cost of moving a session off the victim’s device. It does not make an infected device safe. Malware may still act through the local browser while the user is signed in.
Keep endpoint security, patching, monitoring, and rapid revocation. Review the permissions held by browser extensions on sensitive accounts as well.
Step 1: map your current session design
Before testing DBSC, document how a session starts, refreshes, and ends. Record cookie lifetime, refresh endpoints, logout behaviour, device trust, risk signals, and revocation paths.
Include single sign-on. A user may authenticate with an identity provider and then create sessions across several applications. The standards work is still exploring how binding should travel through those flows.
Draw one diagram. Mark every point where a long-lived token can be copied and used somewhere else. That shows whether device binding addresses the most important risk.
Step 2: choose a narrow pilot
Start with one application and one group of managed devices. Good candidates hold sensitive data, have strong browser management, and can tolerate extra logging. Avoid starting with every customer-facing login.
Define success before the test. Useful measures include registration success, refresh latency, sign-in failures, fallback use, help-desk tickets, and blocked replay attempts.
Keep a control group. Without one, a normal rise in login problems may look like a DBSC issue, or a real issue may hide inside normal noise.
Step 3: design the fallback before launch
Not every browser or device will support the same hardware. A damaged secure module, a browser change, or a recovery event may also break proof. Decide what happens then.
A fallback should not turn into a permanent bypass. It may require fresh authentication, a managed-device check, a shorter session, or limited account access. Log each fallback and review repeated use.
Recovery must work when the original device is lost. Test it with the same care as the happy path. Our guide to emergency admin access shows why every recovery control should be tested by a real person before an incident makes the normal sign-in path fail.
Step 4: keep cookies short and focused
DBSC works by refreshing short-lived cookies after proof. That does not mean every cookie should carry broad authority. Split high-risk actions from ordinary browsing where the application allows it.
Use secure cookie settings, narrow scope, and server-side revocation. Require a fresh check for actions such as changing recovery details, exporting data, creating API keys, or adding an administrator.
Do not let device binding excuse weak session hygiene.
Step 5: watch the user experience
A security control that locks out valid users will be bypassed. Test laptop replacement, sleep and wake, browser restart, profile reset, remote desktop, virtual machines, and device repair. Include accessibility and low-connectivity cases.
Watch the tail. A small delay on many background requests can feel like a broken application.
Give support staff a clear status view without exposing private key material. They need to know whether registration failed, proof failed, or policy blocked the device.
Step 6: plan detection and response
Log session registration, refresh, failed proof, fallback, revocation, and device replacement. Join those events with identity and endpoint alerts.
A burst of copied cookies that fail away from the device is still an incident. DBSC may have stopped the replay, but malware probably reached the original machine. Investigate the endpoint, revoke related sessions, and check for data exposure.
Run a tabletop exercise. CISA provides customizable tabletop exercise packages that teams can adapt to session theft and account recovery.
Questions to ask a vendor
- Which browsers, operating systems, and hardware stores are supported?
- How are session keys created, rotated, and removed?
- What happens during SSO, device replacement, and account recovery?
- Can administrators see proof failures and fallback use?
- Which actions require a fresh sign-in?
- How quickly can one session or every session be revoked?
- Does the design avoid stable device identifiers?
Ask for evidence from a real pilot. A roadmap slide is not a working control.
A safe rollout sequence
- Map the session and threat model.
- Confirm that each supported browser and managed device can register.
- Build registration, refresh, revocation, and recovery.
- Test with managed staff accounts.
- Measure proof failures, recovery steps, and all fallback use.
- Run a stolen-cookie exercise.
- Expand only when support and recovery are ready.
Keep the rollout reversible. Standards and browser behaviour can evolve, especially around federated identity and broader device support.
Frequently asked questions
What is a device-bound session?
It is a signed-in web session that requires proof from a private key stored on the original device before the server refreshes important cookies.
Does DBSC replace passkeys or MFA?
No. Passkeys and MFA protect sign-in. DBSC is designed to reduce abuse after a session cookie is stolen, so the controls work together.
Can DBSC stop malware on the device?
No. Malware acting locally may still use an active session. Device binding mainly makes stolen cookies less useful when moved to another device.
Primary sources
- Google Security Blog: Protecting cookies with DBSC
- W3C Web Application Security: DBSC specification repository
- CISA: Tabletop Exercise Packages