Contractors, partners and staff on personal devices need access to corporate applications, and the two common ways to keep that access away from an unmanaged endpoint, remote browser isolation and virtual desktops, differ by an order of magnitude in cost, scope and operational burden. Remote browser isolation (RBI) runs only the browser remotely and streams a safe rendering to the user. A virtual desktop (VDI, or desktop as a service) runs an entire Windows or Linux session remotely, with every application in it. Choosing the wrong one either leaves a gap (thick client applications that a browser cannot reach) or buys a whole desktop estate to deliver three web apps. This guide compares the two as a decision: what each protects against, how each fails, the user experience and cost drivers, and a worked scenario for third-party access. It is for security architects, end-user computing teams and identity owners. How each technology works is in the Remote Browser Isolation Guide and the Virtual Desktop (VDI) and DaaS Security Guide; this page goes deeper on the choice.
Key takeaways
- The short answer is RBI when every target is a web app, VDI when any target is not. If the user needs a thick client, a database tool, a legacy line-of-business app or local files, a browser-only control cannot deliver it.
- They protect against different things by default. RBI is designed first to keep hostile web content off the endpoint; VDI is designed first to keep applications and data off the endpoint. Both can do some of the other job, but only with deliberate configuration.
- Neither makes the endpoint trustworthy. Keyloggers, screen capture and phone cameras on an unmanaged device still see what the user sees, so identity strength, session controls and watermarking matter in both models.
- VDI is a second estate to run. Session hosts need patching, EDR, application control, egress filtering and admin access controls, which is a large operational cost compared with a browser isolation service.
- Many organisations need both. A common pattern is RBI (or an enterprise browser) for web apps used by most third parties, with VDI reserved for the smaller group that needs full applications.
The short answer and the comparison table
Use RBI when the access need is web-only and the main goal is either to stop web-borne code reaching the endpoint or to stop data being downloaded, copied or printed from web apps. Use VDI when users need non-web applications, persistent desktops, or a controlled environment where the whole working set lives in the data centre or cloud. CISA's guidance on securing web browsers describes isolation as creating a logical barrier between the browser and the operating system on the premise that all web traffic is untrusted, with remote isolation moving processing into a virtualised or cloud container and forwarding cleaned content to the user.
| Dimension | Remote browser isolation | VDI or DaaS |
|---|---|---|
| What runs remotely | A disposable browser session | A full desktop or published applications |
| Application reach | Web and SaaS applications only | Web, thick client and legacy applications |
| Primary protection | Web exploits and malicious downloads never execute on the endpoint | Applications and data stay in the hosting environment |
| Data controls | Policy on upload, download, copy, paste, print and keyboard input per site or category | Clipboard, drive, printer and USB redirection controls, screen capture protection, watermarking |
| Endpoint requirement | Often clientless through a standard browser | A client app or HTML5 client; some features need the native client |
| Persistence | Usually none; session discarded at close | Pooled (non-persistent) or personal (persistent) desktops |
| What you operate | Policy, identity integration and logging; the provider runs the browsers | Images, session hosts, profiles, patching, EDR, application control, capacity |
| Cost drivers | Per-user or per-session licensing; bandwidth | Compute and storage per session, operating system and application licensing, staff time |
| User experience risk | Site breakage, latency, rendering artefacts | Login time, profile issues, latency, peripheral support |
RBI: threat model and how it fails
RBI assumes the web content is the threat. The browser engine runs in a remote container, so a drive-by exploit (MITRE ATT&CK T1189 Drive-by Compromise) lands in a disposable session rather than on the endpoint. CISA lists the benefits as isolating malicious content, reducing reliance on allow and block lists, policies that range from isolating some traffic to isolating every download and link, file transfer controls for webmail and downloads, and logs from the virtual browser.
- Data, not code, is the third-party risk: for contractor access the concern is usually exfiltration, not infection. RBI only helps if upload, download, clipboard and print controls are set for the sanctioned apps; isolation alone still lets a user read and retype data.
- Session theft is not solved: an attacker who steals the user's identity provider session or cookies (T1185 Browser Session Hijacking) can often reach the same SaaS app directly, outside the isolation path. Isolation must be enforced by the application's access policy, not just offered.
- Bypass routes: if the SaaS app can also be reached through the normal internet, or a mobile app, users will drift to whichever path is easier. Restrict sign-in to the isolation path for the user groups in scope.
- Provider trust: the isolation service decrypts and renders the traffic, so it sees the data. Treat it as a high-impact supplier with its own access, logging and data residency review.
- Site breakage: complex web apps, video and some authentication flows render poorly, which leads to exception lists that quietly remove protection.
VDI: threat model and how it fails
VDI assumes the endpoint is the threat and keeps the application and data on hosts the organisation controls. It is also a remote access service exposed to the internet (T1133 External Remote Services) whose session hosts become shared Windows estates. Platform documentation shows what the customer controls (platforms listed alphabetically). Amazon WorkSpaces trusted devices blocks sign-in or reconnection from client devices that cannot present a certificate chaining to one of up to two imported root certificates. Microsoft's Azure Virtual Desktop security recommendations illustrate what the customer side involves: multifactor authentication and Conditional Access for all users, idle and disconnection limits, drive, clipboard, printer and USB redirection disabled by default and enabled only where needed, application control on session hosts, no local admin rights in multi-session pools, no direct RDP to hosts, Trusted Launch with Secure Boot and vTPM, Credential Guard, EDR on every host and egress filtering because session hosts can reach any internet resource by default.
- Redirection undoes isolation: clipboard and drive redirection turned on for convenience moves data straight to the unmanaged device. Use one-way clipboard transfer where the platform supports it.
- Screen and camera capture: screen capture protection blocks software capture on supported clients, but Microsoft states it is not DRM and does not stop someone photographing the screen; watermarking makes leaks traceable.
- Shared host compromise: on multi-session hosts one compromised user session can affect others. Application control, no admin rights and EDR are the controls, and a VDI and DaaS compromise playbook is needed.
- Unfiltered egress: a virtual desktop with open internet access becomes a clean, corporate-IP launch point. Filter egress through a firewall or proxy.
- Weak client trust: without device checks, any machine with valid credentials gets a desktop. Turn on the platform's certificate-based device check and block client types you do not support.
Contractor and BYOD access: matching the tool to the need
| Access need | Better fit | Why |
|---|---|---|
| Contractor using SaaS ticketing, wiki and email | RBI or enterprise browser | All web; download and paste controls give data protection without a desktop |
| Offshore developer needing an IDE, Git and build tools | VDI | Thick clients and source code must stay in the hosted environment |
| Call centre agent on personal device using a CRM web app | RBI with strict data controls | Low cost per user, clientless, fast onboarding |
| Finance user on BYOD needing an ERP thick client | VDI or published application | Browser cannot run the client |
| Administrator managing servers | Neither alone | Use a privileged access workstation or jump host with PAM; see Privileged Access Workstations vs Jump Servers |
| Staff opening uncategorised links from email | RBI | Isolating risky categories is what RBI was designed for |
NIST SP 800-46 Rev. 2 asks that every component of telework and remote access, including organisation-issued and BYOD client devices, be secured against threats identified through threat models. The NCSC's BYOD guidance makes the same point from the other side: how thoroughly the device can be managed determines how well data can be protected, so for an unmanaged device the protection has to sit in the access path. See the BYOD Security Programme guide and the Third-Party Access Guide.
Cost and user experience
- RBI costs: licences per user or session, usually from an SSE or isolation provider, and bandwidth for streaming. CISA notes that isolation may carry the greatest initial cost of its browser recommendations, but may have a lower lifecycle cost through reduced incident response, recovery and filtering maintenance, and suggests treating it as part of a broader architecture change.
- VDI costs: compute and storage sized for peak concurrency, operating system and application licensing, profile storage, image management and the staff to patch, monitor and support a second Windows estate. Pooled desktops cost less than personal ones but limit what users can install.
- Experience with RBI: near-native for most sites; problems show up as broken uploads, single sign-on loops and video quality. Pilot against the real application list.
- Experience with VDI: the user gets a familiar desktop but pays in login time, latency for graphics-heavy work and peripheral quirks; long-running workloads suffer if idle limits are too tight.
- Hidden costs: for RBI, exception lists and supplier review; for VDI, the security tooling every session host needs and the incident response scope of a shared host estate.
When to use each
- Choose RBI when all target apps are web or SaaS, users are numerous and short-term, onboarding speed matters, and the main risks are malicious content or download and copy of data.
- Choose VDI when users need non-web applications, data must never reach the endpoint in any form, the session needs a controlled toolset, or regulation requires processing in a defined environment.
- Choose an enterprise browser instead when devices can take an installed browser and you mainly need data controls, not remote rendering; see the Enterprise Browser and Browser Security Guide.
- Combine when a population splits: RBI for the majority on web apps, VDI for the minority with thick-client needs, both behind the same identity provider and Conditional Access policies.
Small organisations: a managed SaaS estate rarely justifies VDI; RBI included in an SSE subscription, or an enterprise browser, usually covers contractors. Large organisations: expect both, with a decision rule per application recorded in the access catalogue so new third parties are routed consistently.
Worked scenario: an illustrative insurer onboarding 600 contractors
An insurer brings in 600 contractors from two suppliers for an 18-month claims programme, on supplier-owned laptops it cannot manage. The first plan was a pooled virtual desktop for everyone. The application review showed that 540 contractors used only the claims SaaS platform, the document portal and email, all web. Sixty actuarial contractors needed a thick-client modelling tool and access to a data warehouse.
The design that followed put the 540 on RBI with downloads blocked, upload allowed only to the claims platform, clipboard out of the isolated session blocked, printing disabled and watermarking on the document portal. Conditional Access allowed those apps only through the isolation path for the contractor group. The 60 actuaries received pooled virtual desktops with redirection off, application control, EDR, egress limited to the warehouse and approved update services, and screen capture protection on supported clients. Both groups used phishing-resistant MFA and time-bound accounts tied to the supplier contract, following the Contractor Identity Lifecycle Guide. The desktop estate shrank to a tenth of the first plan, and the supplier offboarding at month 18 was a group membership change.
Implementing the decision
- Catalogue applications per population: list every app each user group needs and mark it web, thick client or legacy.
- Define the data rules: per app, decide whether download, upload, copy, print and screen capture are allowed.
- Route by identity: enforce the chosen path through Conditional Access or the SSE policy so users cannot reach in-scope apps directly.
- Pilot with real users: run a pilot ring of 20 to 50 users for two weeks, logging breakage, latency complaints and exception requests.
- Harden VDI before scale: application control, EDR, egress filtering, no admin rights and redirection off on session hosts before the first production user.
- Roll out in waves with rollback: keep the previous access path available to a fallback group, removable by group membership, until each wave is stable.
- Review exceptions monthly: every isolation bypass or redirection exception needs an owner and an expiry.
Detection and measurement
Collect isolation logs (sites, file transfer attempts, blocked actions), VDI connection and session logs, identity provider sign-in logs and session host EDR telemetry in the SIEM. Alert on sign-ins to in-scope apps that did not come through the isolation path, repeated blocked download or paste attempts by one user, VDI sign-ins from untrusted client types and outbound connections from session hosts to unapproved destinations.
| Metric | What it shows | Target direction |
|---|---|---|
| Share of third-party sessions on the approved path | Whether routing is enforced | Close to 100 per cent |
| Active isolation bypass and redirection exceptions | Erosion of the control | Falling, each with owner and expiry |
| Blocked exfiltration actions per 100 users | Pressure on data controls; investigate outliers | Stable or falling |
| Session hosts with EDR, application control and current patches | VDI estate hygiene | 100 per cent |
| Cost per active user per month for each model | Whether the split is still right | Reviewed quarterly |
| Time to revoke a third party's access at contract end | Identity lifecycle working | Same day |
Common failure patterns
- VDI for everyone: a desktop estate built for users who only need web apps, then under-patched because it is too large to run well.
- Isolation offered, not enforced: RBI deployed but the SaaS app still accepts direct sign-ins from the same users.
- Redirection on by default: clipboard and drive redirection enabled in VDI for convenience, turning the desktop into a file transfer service.
- Treating isolation as identity: weak MFA on the isolation or VDI entry point, so stolen credentials reach the protected apps anyway.
- Open egress from session hosts: virtual desktops used as a trusted launch point to the internet and internal networks.
- Unowned exceptions: sites added to the isolation bypass list or users granted redirection without expiry.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev. 5 | AC-17, AC-20 | Remote access authorisation and the use of external, unmanaged systems |
| NIST SP 800-53 Rev. 5 | SC-39, SC-18, SC-7 | Process isolation, mobile code and boundary protection, where RBI and VDI separate code and data from the endpoint |
| NIST CSF 2.0 | PR.AA-03, PR.IR-01, DE.CM-01 | Authenticating users and devices, protecting environments from unauthorised logical access, monitoring network activity |
| MITRE ATT&CK | T1189 with M1048 and M1021 | Drive-by compromise mitigated by application isolation and restricting web content (RBI) |
| MITRE ATT&CK | T1133, T1185, T1078 | Remote services exposure (VDI), session hijacking and valid account abuse that bypass both models without strong identity |
Practitioner checklist
- Classify every application a third-party or BYOD group needs as web, thick client or legacy before choosing a model.
- Write per-application data rules for download, upload, clipboard, print and screen capture.
- Enforce the chosen path in Conditional Access or SSE policy; block direct access for in-scope groups.
- Require phishing-resistant MFA and time-bound accounts on both entry points.
- On VDI, turn redirection off by default, apply application control and EDR, and filter session host egress.
- Use watermarking and screen capture protection where sensitive data is displayed.
- Review the isolation provider as a high-impact supplier: data handling, residency, admin access and logs.
- Track exceptions with owners and expiry dates and review them monthly.
Standards and references
- CISA: Securing Web Browsers and Defending Against Malvertising
- NIST: SP 800-46 Rev. 2 Guide to Enterprise Telework, Remote Access, and BYOD Security
- NIST: SP 800-53 Rev. 5
- NCSC: Device Security Guidance, Bring Your Own Device
- AWS: Restrict WorkSpaces Personal access to trusted devices
- Microsoft Learn: Security recommendations for Azure Virtual Desktop
- Microsoft Learn: Screen capture protection in Azure Virtual Desktop
- MITRE ATT&CK: Drive-by Compromise
Related NHI Mgmt Group resources: Remote Browser Isolation Guide · Virtual Desktop (VDI) and DaaS Security Guide · Third-Party Access Guide · ZTNA Guide