A practical framework for evaluating browser security solutions: defining requirements, testing vendors, and building the business case.
Enterprise work has migrated into the browser, but security hasn’t followed. Attackers have had a lot of success by targeting the browser security gap in the security stack.
If you’re making the case for browser security in your organization, you’ve probably heard the objection: “We already have a SWG, EDR, email phishing tools, an IdP — can’t we solve this with what we have?” But modern attacks route past every one of them. Step through each control layer to see where its coverage ends.
The market has converged on “Secure Enterprise Browsers” as a single category term, but it buckets together fundamentally different solutions serving different buyers with different problems.
Deploys into the browser & ready to use
Architecture ADeploy into the browsers users already have without migration, retraining, or workflow changes. Collect telemetry from inside the browser session, detect malicious activity, and enforce controls. Gartner notes that extensions have become the preferred form factor, citing easier deployment, minimal training, and multi-browser compatibility, with extension-based offerings making up the majority of SEB vendors today.
The whole browser is replaced
Architecture BReplace the user's browser with a controlled workspace platform. Optimized for IT teams solving for access governance, device posture, VDI replacement, and policy enforcement. The Achilles heel is migration: adopting one means asking every user to switch away from Chrome, and Gartner notes most buyers start with a single use case and rarely pursue organization-wide deployment.
Use cases ranked by risk reduction delivered and the browser’s unique suitability as the control point. Tier 1 use cases are uniquely solved in the browser.
High security value · no other layer can see the attack
Account takeover is the dominant entry point for enterprise breaches. The attack surface extends well beyond what identity tooling can see: ghost logins, shadow SaaS apps outside IdP coverage, accounts without MFA, and credentials already compromised in earlier breaches.
Account takeover is a browser-native attack vector — attackers target users inside their browsers to compromise their identities on cloud services. The browser is the only layer with visibility of the full identity attack surface and the TTPs used to exploit it.
AiTM phishing — where a reverse proxy intercepts credentials and session tokens in real time — is now the standard technique for bypassing MFA at scale. 1 in 2 payloads are now delivered outside email, and most attacks use redirects and cloaking techniques to mask malicious pages. The threat extends beyond credential phishing: authorization attacks like device code phishing bypass the login altogether (including passkeys).
Modern phishing pages are dynamic web applications rendered by the browser's JavaScript engine. Email gateways never see the destination page, and proxies can't see what's happening client-side. Blocklists chase the most disposable indicators attackers have. The browser is the only layer where you can observe the rendered page as the user sees it — the same shift the industry made from signature-based AV to behavioral EDR.
The average employee maintains ~15 identities across ~220 applications, and fewer than 1 in 3 support SAML SSO. In one snapshot, we found that 37% of accounts lacked MFA entirely and roughly 9% had a breached, weak, or reused password. Ghost logins — legacy local credentials that persist alongside SSO — appear in roughly 10% of accounts, providing attackers with a way around enterprise SSO.
IdPs enforce policy for federated applications, but self-adopted, unmanaged applications sit outside the reach of the IdP. The browser sees the login itself, discovering the identity at the moment it's created or used: authentication method, password strength, MFA status, and all. Being in the browser means acting at the moment of authentication — not just building an inventory.
The Cyberhaven campaign compromised 35+ Chrome extensions reaching ~2.6 million users. DarkSpectre compromised 8.8 million browsers through extensions that held "verified" web store status throughout a seven-year operation. Analysis of 20,000+ enterprise extensions found 46.76% hold the permission combinations needed to perform account takeover with no user interaction.
SWGs and network tools are structurally blind to extensions running inside the browser: a malicious extension exfiltrating session tokens generates no anomalous network signal. Endpoint agents have no visibility into extension behavior at the session level. Even the stores' own review consistently fails to catch malicious updates. If detection can't happen at the store level, it has to happen in the browser.
45% of applications are purchased through expense channels rather than IT procurement, and that misses the larger population of self-adopted freemium accounts. OAuth connections across these unmanaged estates have become devastating: the ShinyHunters Salesforce campaign pivoted through harvested OAuth tokens to reach hundreds of connected services. Once issued, a token survives password resets, MFA changes, and session revocations.
Every OAuth consent grant transits the browser: the authorization prompt, the scope disclosure, the approval click, and the redirect. That makes the browser the only layer where an unwanted grant can be intercepted before the token is issued — and the only layer that observes the full application estate, because it captures the login itself at the moment an employee authenticates.
Includes “what good looks like” criteria, vendor questions, and how to test it in a PoC.
High security value · works alongside other layers
ClickFix was the most common initial access vector reported by Microsoft in 2025. The technique writes a malicious command to the clipboard and social-engineers users into executing it — fileless, user-executed, with 4 in 5 payloads arriving via search engines rather than email. New sub-techniques, lures, payloads, and ways of executing the malicious commands are constantly being added to ClickFix kits.
The browser is the earliest intervention point: detecting the clipboard injection and social engineering lure before anything reaches the endpoint in executable form. For ConsentFix, the browser is the only detection layer. When attacks arrive via malvertising, compromised websites, or trusted platform domains, the page loading in the browser is the first and often only detection opportunity.
AI adoption is outpacing security governance at nearly every organization. But the security problems AI creates are largely existing Tier 1 problems amplified by a new category of tooling: shadow AI apps are shadow SaaS, AI OAuth integrations are OAuth governance, AI browser extensions are extension security, and personal AI account usage is an identity posture problem.
The browser sees the apps, the OAuth grants, the extensions, and whether the user is authenticated via a corporate or personal account. For unsanctioned AI tools, it provides the only real-time enforcement point. For AI agents operating through OAuth grants, the browser is where the consent decisions that authorize them are made.
Endpoint logs show what processes executed, network logs show traffic destinations, IdP logs show authentication events — but none show what happened inside the browser session. This is the missing middle of modern investigations: the link between "a user visited a URL" and "credentials were submitted to a phishing page that issued a session token now being replayed."
The browser fills the session gap that other sources miss. User and session timelines reconstruct the full sequence: what users clicked, what pages loaded, what credentials were entered, and what session activity followed. The same telemetry supports proactive scenarios like blast radius assessment when a third-party vendor announces a breach.
Infostealers are the upstream supply chain for a disproportionate share of the most damaging attacks — harvesting credentials, session cookies, and browser profile data en masse. The chain is self-reinforcing: ClickFix delivers infostealers, infostealers harvest accounts, compromised accounts distribute more ClickFix lures.
The browser is relevant at three points in the kill chain: delivery interception (ClickFix is the primary delivery mechanism), detecting stolen factors in use (breach dataset matching at login, session marker injection for stolen tokens), and as the credential leakage path itself (personal-account browser sync copies corporate passwords to unmanaged personal cloud accounts).
File uploads to unsanctioned services, sensitive data pasted into AI tools, and exfiltration through personal accounts are genuine risks. But DLP is a multi-layer problem: the browser is an important layer but not a standalone replacement — it can't see email-based loss, endpoint-to-endpoint transfers, or cloud API exfiltration.
Browser-layer controls provide real value, particularly for BYOD and contractors where endpoint DLP agents can't be deployed. The browser sees file uploads, downloads, clipboard activity, and AI prompt inputs, and can intervene in real time to stop risky and unapproved actions.
Includes “what good looks like” criteria, vendor questions, and how to test it in a PoC.
Lower browser-specific value · evaluate last
Blocking known-bad URLs and domains is a long-established control, and it still removes commodity threats cheaply. But attackers rotate infrastructure faster than any list can track — most phishing domains are live for under two days, behind bot protection — so reputation-based blocking only catches what has already been seen elsewhere.
Secure web gateways, DNS filtering, and SASE platforms already enforce URL and domain policy across every app and device, not just the browser. A browser solution can apply the same lists, but it adds little a network control doesn't already provide — evaluate it as a convenience, not a reason to buy (and URL blocking doesn't mean effective phishing protection).
As AI assistants and agentic browsers read web content on a user's behalf, instructions hidden in a page, document, or email can redirect what the model does — leaking data or taking actions the user never asked for. It's an emerging risk, and the techniques are still changing quickly.
Prompt injection is ultimately a weakness in how the model and the AI application handle untrusted input, so the strongest controls sit with the AI provider and the application's own guardrails. The browser adds visibility into which AI tools are in use and what is pasted into them, but it can't reliably tell a malicious instruction from ordinary page content.
Exploits against the browser engine itself — memory corruption and zero-days used to escape the sandbox — are real but rare, and mostly reserved for high-value targets. Modern browsers patch quickly and update automatically, which keeps the window for most exploits short.
Browser vendors and endpoint protection are better placed here: vendors fix the engine, and EDR sees what an exploit does to the operating system once it escapes the sandbox. Keeping browsers up to date delivers most of the value.
Includes “what good looks like” criteria, vendor questions, and how to test it in a PoC.
How easily each solution deploys into your environment, against how much security value it delivers on the threats the evidence says matter most.
Select a solution to explore
Built for the Tier 1 threat model: identity attacks inside the browser session. Technique-class behavioral detection that fires even on previously uncatalogued kits on fresh infrastructure, backed by a dedicated threat research team.
Extension-based deployment in minutes across Chrome, Edge, Firefox, and Brave — no migration barrier — with security value concentrated on the highest-priority use cases in the framework. That combination is what puts it top-right.
What it’s built for: Acquired extension covering the fundamentals: phishing, extensions, AI, and DLP.
Deploys in minutes with no migration barrier. Lands below Push on value: lightweight coverage across use cases rather than deep detection of Tier 1 identity attacks. Expect priorities to shift post-acquisition toward Apple ecosystems.
What it’s built for: Exploit prevention heritage — a Tier 3 use case — with identity claims added post-acquisition. Detection depth on identity attacks is untested at scale.
The extension deploys via JavaScript runtime injection — a stability risk worth testing alongside the Falcon sensor. Value increasingly depends on the Falcon ecosystem; standalone value is unclear.
What it’s built for: Governance and policy orientation — strongest on AI usage control, risk scoring, and access policy rather than detection oriented around attacker TTPs.
Lightweight extension deployment scores well on ease. Since the acquisition, innovation is being redirected to Akamai platform capabilities, with long-term scope shaped by Akamai’s priorities.
What it’s built for: Malware sandboxing heritage: file-focused isolation in an increasingly malware-free attack landscape. No behavioral detection of identity attacks.
Extension-based and easy to deploy, but browser-native innovation is being curtailed to fit Zscaler’s SWG model — the product won’t cannibalize proxy revenue.
What it’s built for: Full browser replacement optimized for IT workspace control: access governance, device posture, VDI replacement, and policy enforcement. Detection is known-bad blocking — policy rules on URLs, domains, and file types, with no behavioral detection.
Adopting Island means asking every user to switch away from Chrome, which keeps deployments scoped to specific user groups. Security value is concentrated in IT controls rather than the attacks driving the most damaging breaches.
What it’s built for: Palo Alto’s full browser, built from the Talon acquisition. SASE-derived heritage — URL filtering and sandboxing — with “Live Page Scanning” layered on WildFire. Value is tied to being a PANW platform customer.
Full browser migration limits coverage, with the extension positioned as a transitional on-ramp. Engineering focus has shifted from browser-native innovation to platform integration since the acquisition.
Chart axes: Security Value rises upward,Deployment Ease rises to the right
Full-stack enterprise browsers sit on the left: they require browser migration that limits deployment scope, and their security value is concentrated in IT workspace controls rather than the Tier 1 identity attacks that drive the most damaging breaches. The acquired extension vendors land in the center: they deploy more easily, but have trade-offs to consider around new priorities post-acquisition and vendor lock-in.
Push lands top-right: extension-based deployment with no migration barrier, and security value concentrated on the Tier 1 identity attacks that the guide’s framework identifies as the highest priority. That positioning follows directly from applying the evaluation criteria the preceding sections describe.
Each dimension maps directly to the use case ranking, evaluation criteria, and architectural analysis above. The framework is the same regardless of vendor — the outcomes differ because of what each was built to do.
| Solutions | Threat Model Alignment | Capability Depth | Deployment & Coverage | Operational Maturity | Innovation Velocity | Strategic Independence |
|---|---|---|---|---|---|---|
| Full-Stack Browsers | ||||||
| Island | IT workspace control. Policy enforcement and known-bad blocking — not built for identity attack detection. | Known-bad blocking. Policy rules on URLs, domains, file types. No behavioral detection. | Full browser replacement; extension is an on-ramp. Migration resistance limits scope. | Mature platform, named enterprise customers. Strong IT workflow integration. | Rapid expansion, but IT/productivity-focused — not security research. | Independent, but full replacement creates migration lock-in. |
| Prisma Browser, Palo Alto Networks | SASE-derived: URL filtering and sandboxing heritage. Identity claims tied to the platform stack. | “Live Page Scanning” layered on WildFire + URL filtering — supplements the reputation model. | Full browser, “transitional” extension, mobile app. Extension positioned to upsell. | Mature via the PANW stack. Value tied to being a PANW customer. | Repositioned since Talon (Dec 2023). Effort shifted to platform integration. | Tied to Prisma SASE. Platform lock-in by design. |
| Acquired Extensions | ||||||
| Falcon Seraphic Enterprise Browser, CrowdStrike | Exploit prevention heritage (Tier 3). Identity claims added post-acquisition. | DOM attribute analysis for phishing (claimed). Depth untested at scale. | Extension via JS runtime injection. Stability risk; test alongside the Falcon sensor. | Falcon integration roadmap not fully shipped — buying future state. | Post-outage stability priority absorbs engineering. Narrowing scope. | Falcon ecosystem lock-in. Standalone value unclear. |
| Zscaler Zero Trust Browser, Zscaler (formerly SquareX) | Malware sandboxing heritage. File-focused isolation in a malware-free landscape. | File sandboxing. No behavioral detection of identity attacks. | Extension-based. Limited pre-acquisition deployment history. | Early-stage ZTE integration. $20M raised, acquired under a year post-Series A. | Curtailed to fit Zscaler’s SWG model — browser-native innovation deprioritized. | Tied to Zscaler platform. Will not cannibalize SWG revenue. |
| Akamai Workforce Protector, Akamai (formerly LayerX) | Governance and policy orientation. Strongest on AI usage control, not threat detection. | AI governance, risk scoring, access policy. Detection oriented around governance, not attacker TTPs. | Extension-based. Lightweight deployment. | Rebranded to Akamai SIA. Integration with Guardicore and EAA underway. | Innovation redirected to Akamai platform capabilities. | Tied to Akamai platform. Long-term scope shaped by Akamai’s priorities. |
| Keep Aware, Jamf | Covers fundamentals: phishing, extensions, AI, DLP. Breadth over depth at seed stage. | Broad but shallow use case coverage. | Extension-based. Deploys in minutes. | Early stage, recently acquired. Investigate whether output supports investigation workflows or generates policy alerts. | Active but constrained by resources. Test against novel kits on fresh infrastructure. | Recently acquired. Tied to Jamf's Apple-centric priorities. |
| Independent Extensions | ||||||
| Push Security | Built for the Tier 1 threat model: identity attacks inside the browser session. ATO, phishing, identity posture, extensions, shadow SaaS, OAuth. | Technique-class behavioral detection. Fires on zero-day kits on fresh infrastructure. Agentic detection pipeline with autonomous threat hunting. | Extension-based. Deploys in minutes across Chrome, Edge, Firefox, Brave — including agentic browsers. No migration required. | Production-scale. SIEM, SOAR, and IdP integration. Investigation timelines and custom detection rules. | Dedicated threat research team. Has broken and named multiple browser-native attack techniques (ConsentFix, InstallFix, cross-IdP impersonation, ghost logins). | Independent. Engineering concentrated entirely on browser security capability. |
Get the full guide for the evaluation criteria, PoC guidance, vendor deep-dives, the complete business case framework, and 50+ questions to ask every vendor you evaluate.
