The VPN Gatekeeper’s DilemmaWhy VPNs are Always in the Crosshairs, what three years of VPN CVEs reveal about enterprise trust, and how ZTNA changes the equation.Every piece of software ships with vulnerabilities, and that includes the security products built to prevent those vulnerabilities from being exploited. The question for any security team is not whether vulnerabilities will surface, but what happens when they do. How much of the environment is exposed, how quickly can an attacker move, and how wide is the blast radius? By these measures, VPN and firewall gateways have a clear structural problem.Since January 2024, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added more than 50 vulnerabilities across VPN and firewall products from major vendors, to its Known Exploited Vulnerabilities (KEV) catalog. That’s roughly one every two to three weeks, and 22 of these have been flagged by CISA as known to have been used in ransomware campaigns. Each came with its own advisory, emergency patch window, and incident-response scramble.Exploited VPN and Firewall flaws from major vendors: Vulnerabilities in VPN and firewall gateway products added to CISA’s Known Exploited Vulnerabilities (KEV) catalog.Taken one at a time, these vulnerabilities look like a string of bad luck for legacy Firewall and VPN providers. Together, they highlight something more fundamental, that traditional VPN architecture puts an enterprise’s trust decisions on the most exposed piece of infrastructure it owns.All of these vendors employ talented engineers who run sophisticated security programs to identify, disclose, and patch vulnerabilities responsibly. When capable teams across an entire product category keep producing the same classes of critical flaws, that’s precisely the point. The problem isn’t any single team, it’s the design flaw in the architecture they’re all working with. The Same Failure Modes Keep RepeatingRemoving product and company names makes it abundantly clear that the VPN vulnerabilities attackers actually exploit follow a small number of recurring patterns. Five failure modes account for nearly all of it.CVEAttack VectorWhat the attacker gotCISA KEVCVE-2018-13379SSL VPN path traversal exposing a session filePlaintext VPN credentials; 87,000 devices’ credentialsNov 2021CVE-2022-42475Pre-auth heap overflow in SSL VPN daemonRCE; roughly 20,000 FortiGates compromisedDec 2022CVE-2023-4966Memory over-read on GatewayLive session tokens that bypass MFAOct 2023CVE-2023-46805 CVE-2024-21887Auth bypass chained with command injectionPre-auth RCE; CISA ordered federal factory resetsJan 2024CVE-2024-21762Pre-auth out-of-bounds write in SSL VPNRemote code executionFeb 2024CVE-2024-3400Pre-auth command injectionRoot on the firewallApr 2024CVE-2024-24919Arbitrary file read on GatewayCredential and password-hash disclosureMay 2024CVE-2025-0282Pre-auth stack overflowRCE plus malware built to survive upgradesJan 2025CVE-2024-53704SSL VPN authentication bypassHijacked active VPN sessionsFeb 2025CVE-2025-5777Memory over-read on GatewayLive session tokens, againJul 2025CVE-2025-20333 CVE-2025-20362Buffer overflow chained with missing authorizationCode execution; bootkit that survives firmware upgradesSep 2025CVE-2026-24858SSO authentication bypassAdmin access to other customers’ devices; zero-dayJan 2026CVE-2026-0257Forgeable authentication override cookieAuthenticated VPN session as any user, no password or MFAMay 2026CVE-2026-0300Pre-auth buffer overflow in Authentication PortalRoot-level remote code executionMay 2026CVE-2026-50751IKEv1 remote access VPN authentication bypassVPN session without a valid passwordJun 20261. Pre-authentication code execution. A VPN concentrator has to accept connections from anyone on the internet before it knows who they are. That listening service parses untrusted input from anonymous senders, which makes it the single most valuable target in the environment. Many of the CVEs listed above are variations of this pattern. A maliciously crafted request reaches code running with high privilege on a device that sits directly between the internet and an enterprise’s internal network.2. Skipped authentication. Attackers obtain or forge the artifact that represents an authenticated session, so the login process never happens. CitrixBleed and CitrixBleed 2 leaked live session tokens from appliance memory. SonicWall vulnerabilities let attackers hijack active SSL VPN sessions. A Check Point exploit granted VPN sessions without a valid password. Palo Alto Networks’ CVE-2026-0257 is the most recent example of the pattern, where GlobalProtect offers an “authentication override” option that issues returning users an encrypted cookie so they can skip the login prompt. Rapid7 found that when the certificate used to encrypt those cookies was also the one presented by the portal’s HTTPS service, an attacker could pull the public key from the TLS handshake and forge a valid cookie for any user. The result was a fully authenticated VPN session with no username, password, or MFA prompt.3. Credential exposure. Because VPN appliances terminate user authentication, they handle credentials, hashes, and session secrets, and file-read or traversal bugs can turn into credential theft. Fortinet exposed plaintext VPN credentials that eventually led to a public dump covering 87,000 devices. Stolen VPN credentials may also remain valid after the appliance is patched unless every affected password is reset.4. Patch-resistant compromises. Attackers who reach an edge appliances increasingly build persistence into their exploits, so they’re designed to survive a defender’s response:In 2024, CISA’s Emergency Directive 24-01 required federal agencies to factory-reset compromised Ivanti appliances. A follow-on joint advisory warned that a sophisticated actor might retain root-level persistence even after a reset.In 2025, Fortinet disclosed that attackers had planted a symbolic link on previously compromised FortiGates. The link preserved read-only access to the device’s file system even after the original vulnerabilities were patched.The UK’s National Cyber Security Centre (NCSC) found that the bootkit used against Cisco ASA devices in 2025 “survives reboots and firmware upgrades.” That campaign prompted CISA’s Emergency Directive 25-03.Mandiant found Ivanti malware that displayed a fake upgrade progress bar while silently blocking the real one.5. Network-level access. Each of these patterns ends the same way. A VPN’s job is to place a remote user on a corporate network. But once an attacker is through the gate, the architecture hands them an internal IP address and routes to everything that address can reach. That’s why so many of these CVEs carry CISA’s “known ransomware campaign use” flag. The real harm comes from lateral movement that the architecture makes possible. Patching Won’t Close the GapThe standard response to new VPN CVEs is correct: patch software, rotate certificates and credentials, and hunt for indicators of compromise. But the industry data shows that defenders are losing the race.Attackers move in days, where defenders take weeks: Key timing benchmarks (in days), relative to a vendor’s patch release.Exploitation comes first. Mandiant’s M-Trends 2026 reports that “exploits” were the top initial attack vector for the sixth year in a row. It also found that the mean time to exploit is now negative, which means that on average, vulnerabilities are being exploited before a patch is available.Edge devices are where exploits happen. The 2025 Verizon Data Breach Investigations Report (DBIR) found that edge devices and VPNs grew to 22% of vulnerability exploitation targets (from 3%) in a single year. Google’s Threat Intelligence Group also counted 21 zero-day exploits in security and networking products in 2025, which was about half of all enterprise-targeted zero-day vulnerabilities it tracked.Vulnerabilities are now the leading way in. The 2026 DBIR found that 31% of breaches now begin with vulnerability exploitation, overtaking credential abuse as the top initial access vector for the first time in the report’s 19-year history.Remediation is getting slower. The same report found that only 26% of KEV-listed vulnerabilities were fully resolved, with a median remediation time of 43 days (up from 32 the year before).Organizations see the gap themselves. In the Zscaler ThreatLabz 2026 VPN Risk Report, 54% of organizations said they needed a week or more to patch a critical VPN vulnerability.Forty-three days, or even a week is a relatively long time when exploitation begins in a matter of days (if not sooner). This gap persists even in well-run security programs because the process, and decisions justifying that process is critical:Appliances must be scheduled for maintenance.Remote access can’t go down during business hours.High-availability pairs have to be upgraded in sequence.Fixes have to be tested against a configuration that has developed years of exceptions.Government guidance on patching vulnerabilities also reflects this operational reality. The Five Eyes intelligence partners issued a joint series on edge-device security in February 2025. CISA’s Binding Operational Directive 26-02 now orders federal agencies to replace edge devices that have reached end of support. In 2024, CISA, the FBI, and international partners jointly urged organizations to move away from traditional VPNs toward zero trust and secure access service edge (SASE) architectures. The Identity Stack Can’t Close the GapOver the past decade, most enterprises have invested heavily in identity-based solutions like single sign-on, multi-factor authentication (MFA), conditional access, and device posture checks. Those investments help organizations keep unauthorized users out of enterprise networks. In fact, Microsoft’s 2025 Digital Defense Report states that MFA can block over 99% of identity-based attacks.When we take a closer look at the failure modes attackers are utilizing, however, it’s clear that identity based solutions have a limited impact. Pre-authentication exploits never reach the identity provider. As far as the appliance is concerned, a forged or stolen session token is presented after authentication has already happened. Leaked credentials can be reused against a gateway that doesn’t enforce phishing resistant MFA. In each case the identity stack did its job, but the architecture offered the attacker a path that never touched it.This is the gatekeepers dilemma; a convenient way for legitimate users to access a network is also the most likely attack point, which is why convenience and ingress/egress features keep appearing in these security advisories. Persistent sessions, authentication override cookies, and single sign-on shortcuts exist because VPN reauthentication is painful for users. In a VPN architecture every convenience becomes a trust artifact stored or validated at the internet-facing edge. Administrators are forced to choose between a worse user experience and a larger more vulnerable attack surface.The consequences of that choice are visible in the ransomware data that Zscaler and other organizations have collected over the past decade. The Zscaler ThreatLabz 2026 Ransomware Report found that data exfiltration among the ten most active ransomware groups surged 275.8% year over year, to 896.2 terabytes. It’s Time to Change Where the Architecture SitsZero trust network access (ZTNA) doesn’t make software vulnerability-free, and no honest vendor should claim that it does. Every access platform, including its cloud brokers and endpoint clients, is software that needs secure engineering and rapid patching. But ZTNA changes the architecture those vulnerabilities exist in, and that changes everything.Where the trust decision lives: A VPN exposes a gateway and hands out network access. ZTNA exposes nothing and brokers access to one app at a time.Four principles separate these two architectures.Applications are invisible to the internet. In a cloud-native ZTNA model, private applications and the lightweight connectors in front of them make only outbound connections to a cloud exchange. Because nothing listens for inbound connections or exposes public IPs, applications are rendered completely invisible to the internet—leaving no portal for an attacker to discover, fingerprint, scan, or exploit.Access is segmented to specific applications, not the network. Rather than extending subnets or placing users on routed segments, a Zero Trust session connects authorized users, workloads, and AI agents directly to designated applications. By dynamically discovering private workloads and automating least-privilege user-to-app segmentation, lateral threat movement is eliminated. Even if a session is compromised, an attacker never receives an internal IP address or network routes.Trust is continuously verified and reinforced with active containment. Instead of relying on a forgeable bearer token or override cookie cached at an internet-facing gateway, identity, context, and device posture are evaluated continuously in the cloud throughout the session. Users get a seamless experience without perimeter shortcuts—and by integrating deception decoys directly into the access layer, organizations can immediately detect and isolate any post-compromise reconnaissance.Patching moves to the cloud. When the access layer is a multi-tenant cloud service, platform updates are deployed seamlessly without waiting for customer maintenance windows or complex HA upgrades. Modern Zero Trust architectures can also continuously scan private applications for specific exploit paths and apply precise inline virtual patches at machine speed, closing the 43-day remediation gap long before vulnerabilities can be weaponized.Here’s how each failure mode plays out under these two models:Failure modeTraditional VPNZero trust architecturePre-auth exploitationConcentrator listens for inbound connections on the public internetApplications and connectors make outbound-only connectionsSession bypassSession state is held and validated at the exposed edgeSessions are brokered per user and per application in the cloud, with continuous identity and posture checksCredential exposureCredentials and session secrets are handled on the applianceAuthentication is delegated to the identity provider; the access layer never stores user credentialsPersistence after patchingCustomer-operated appliance; attacker can hide in firmwareCloud-operated service with no customer-patched gateway to embed malwareNetwork-level accessSuccessful entry places the attacker on the networkUsers only connect to specific applications  Solving the Gatekeepers DilemmaCritical vulnerabilities like PAN-OS (CVE-2026-0257), FortiOS (CVE-2024-21762), PHP2Shell (CVE-2026-109618) and others are now endemic, and this trend will likely accelerate due to frontier AI-powered attacks. In addition to data loss and system compromise, CVEs take an enormous toll on organizations in the form of monitoring and patching complex global systems.Statistically, the next critical vulnerability is just a few weeks away. When it arrives, the response playbook will be the all-too-familiar emergency patching, certificate and credential rotation, log reviews, and a lingering concern that the attacker may have left something behind.Organizations can keep running that playbook indefinitely, or they can recognize that the playbook exists because the architecture demands it, and change the architecture.Administrators who enabled the features exploited in recent advisories weren’t being careless, they were trying to solve real user problems within the limits of a legacy architecture. But when dozens of VPN and firewall vulnerabilities have landed on CISA’s KEV catalog in just the last three years, it’s clear that organizations can’t patch their way out of an architectural design flaw. Zero trust solves the gatekeeper’s dilemma by removing exposed gateways altogether, to give users seamless and direct access to applications without exposing your corporate network to the internet.Zscaler Private Access (ZPA), the industry’s first and only AI-powered ZTNA solution, eliminates the gatekeeper’s dilemma by connecting authorized users directly to specific private applications through the Zscaler Zero Trust Exchange. Instead of defending an exposed perimeter gateway that leads to your corporate network, ZPA secures private applications across four core pillars:Hide: Applications are not exposed to the internet. Lightweight App Connectors establish inside-out connections, eliminating public IPs and exposed DMZ gateways so threat actors have nothing to scan, fingerprint, or attack.Segment: Access is brokered on a least-privileged, per-user and per-application basis, with continuous identity and device posture checks. AI-powered app segmentation dynamically recommends and enforces granular user-to-app policies that eliminate lateral movement.Shield: Real-time inline threat and data protection blocks exploits in transit, protecting and patching vulnerable private applications before attackers can strike.Deceive: Native deception and decoys lure, detect, and isolate sophisticated threat actors and malicious insiders attempting post-compromise reconnaissance.Because ZPA is delivered as a cloud-native service across 200+ locations, there is no VPN waiting for an emergency maintenance window when the next critical CVE surfaces.Ready to leave emergency VPN patching behind?Schedule a demo to see Zscaler in actionRead the ThreatLabz 2026 VPN Risk Report  

​[#item_full_content] The VPN Gatekeeper’s DilemmaWhy VPNs are Always in the Crosshairs, what three years of VPN CVEs reveal about enterprise trust, and how ZTNA changes the equation.Every piece of software ships with vulnerabilities, and that includes the security products built to prevent those vulnerabilities from being exploited. The question for any security team is not whether vulnerabilities will surface, but what happens when they do. How much of the environment is exposed, how quickly can an attacker move, and how wide is the blast radius? By these measures, VPN and firewall gateways have a clear structural problem.Since January 2024, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added more than 50 vulnerabilities across VPN and firewall products from major vendors, to its Known Exploited Vulnerabilities (KEV) catalog. That’s roughly one every two to three weeks, and 22 of these have been flagged by CISA as known to have been used in ransomware campaigns. Each came with its own advisory, emergency patch window, and incident-response scramble.Exploited VPN and Firewall flaws from major vendors: Vulnerabilities in VPN and firewall gateway products added to CISA’s Known Exploited Vulnerabilities (KEV) catalog.Taken one at a time, these vulnerabilities look like a string of bad luck for legacy Firewall and VPN providers. Together, they highlight something more fundamental, that traditional VPN architecture puts an enterprise’s trust decisions on the most exposed piece of infrastructure it owns.All of these vendors employ talented engineers who run sophisticated security programs to identify, disclose, and patch vulnerabilities responsibly. When capable teams across an entire product category keep producing the same classes of critical flaws, that’s precisely the point. The problem isn’t any single team, it’s the design flaw in the architecture they’re all working with. The Same Failure Modes Keep RepeatingRemoving product and company names makes it abundantly clear that the VPN vulnerabilities attackers actually exploit follow a small number of recurring patterns. Five failure modes account for nearly all of it.CVEAttack VectorWhat the attacker gotCISA KEVCVE-2018-13379SSL VPN path traversal exposing a session filePlaintext VPN credentials; 87,000 devices’ credentialsNov 2021CVE-2022-42475Pre-auth heap overflow in SSL VPN daemonRCE; roughly 20,000 FortiGates compromisedDec 2022CVE-2023-4966Memory over-read on GatewayLive session tokens that bypass MFAOct 2023CVE-2023-46805 CVE-2024-21887Auth bypass chained with command injectionPre-auth RCE; CISA ordered federal factory resetsJan 2024CVE-2024-21762Pre-auth out-of-bounds write in SSL VPNRemote code executionFeb 2024CVE-2024-3400Pre-auth command injectionRoot on the firewallApr 2024CVE-2024-24919Arbitrary file read on GatewayCredential and password-hash disclosureMay 2024CVE-2025-0282Pre-auth stack overflowRCE plus malware built to survive upgradesJan 2025CVE-2024-53704SSL VPN authentication bypassHijacked active VPN sessionsFeb 2025CVE-2025-5777Memory over-read on GatewayLive session tokens, againJul 2025CVE-2025-20333 CVE-2025-20362Buffer overflow chained with missing authorizationCode execution; bootkit that survives firmware upgradesSep 2025CVE-2026-24858SSO authentication bypassAdmin access to other customers’ devices; zero-dayJan 2026CVE-2026-0257Forgeable authentication override cookieAuthenticated VPN session as any user, no password or MFAMay 2026CVE-2026-0300Pre-auth buffer overflow in Authentication PortalRoot-level remote code executionMay 2026CVE-2026-50751IKEv1 remote access VPN authentication bypassVPN session without a valid passwordJun 20261. Pre-authentication code execution. A VPN concentrator has to accept connections from anyone on the internet before it knows who they are. That listening service parses untrusted input from anonymous senders, which makes it the single most valuable target in the environment. Many of the CVEs listed above are variations of this pattern. A maliciously crafted request reaches code running with high privilege on a device that sits directly between the internet and an enterprise’s internal network.2. Skipped authentication. Attackers obtain or forge the artifact that represents an authenticated session, so the login process never happens. CitrixBleed and CitrixBleed 2 leaked live session tokens from appliance memory. SonicWall vulnerabilities let attackers hijack active SSL VPN sessions. A Check Point exploit granted VPN sessions without a valid password. Palo Alto Networks’ CVE-2026-0257 is the most recent example of the pattern, where GlobalProtect offers an “authentication override” option that issues returning users an encrypted cookie so they can skip the login prompt. Rapid7 found that when the certificate used to encrypt those cookies was also the one presented by the portal’s HTTPS service, an attacker could pull the public key from the TLS handshake and forge a valid cookie for any user. The result was a fully authenticated VPN session with no username, password, or MFA prompt.3. Credential exposure. Because VPN appliances terminate user authentication, they handle credentials, hashes, and session secrets, and file-read or traversal bugs can turn into credential theft. Fortinet exposed plaintext VPN credentials that eventually led to a public dump covering 87,000 devices. Stolen VPN credentials may also remain valid after the appliance is patched unless every affected password is reset.4. Patch-resistant compromises. Attackers who reach an edge appliances increasingly build persistence into their exploits, so they’re designed to survive a defender’s response:In 2024, CISA’s Emergency Directive 24-01 required federal agencies to factory-reset compromised Ivanti appliances. A follow-on joint advisory warned that a sophisticated actor might retain root-level persistence even after a reset.In 2025, Fortinet disclosed that attackers had planted a symbolic link on previously compromised FortiGates. The link preserved read-only access to the device’s file system even after the original vulnerabilities were patched.The UK’s National Cyber Security Centre (NCSC) found that the bootkit used against Cisco ASA devices in 2025 “survives reboots and firmware upgrades.” That campaign prompted CISA’s Emergency Directive 25-03.Mandiant found Ivanti malware that displayed a fake upgrade progress bar while silently blocking the real one.5. Network-level access. Each of these patterns ends the same way. A VPN’s job is to place a remote user on a corporate network. But once an attacker is through the gate, the architecture hands them an internal IP address and routes to everything that address can reach. That’s why so many of these CVEs carry CISA’s “known ransomware campaign use” flag. The real harm comes from lateral movement that the architecture makes possible. Patching Won’t Close the GapThe standard response to new VPN CVEs is correct: patch software, rotate certificates and credentials, and hunt for indicators of compromise. But the industry data shows that defenders are losing the race.Attackers move in days, where defenders take weeks: Key timing benchmarks (in days), relative to a vendor’s patch release.Exploitation comes first. Mandiant’s M-Trends 2026 reports that “exploits” were the top initial attack vector for the sixth year in a row. It also found that the mean time to exploit is now negative, which means that on average, vulnerabilities are being exploited before a patch is available.Edge devices are where exploits happen. The 2025 Verizon Data Breach Investigations Report (DBIR) found that edge devices and VPNs grew to 22% of vulnerability exploitation targets (from 3%) in a single year. Google’s Threat Intelligence Group also counted 21 zero-day exploits in security and networking products in 2025, which was about half of all enterprise-targeted zero-day vulnerabilities it tracked.Vulnerabilities are now the leading way in. The 2026 DBIR found that 31% of breaches now begin with vulnerability exploitation, overtaking credential abuse as the top initial access vector for the first time in the report’s 19-year history.Remediation is getting slower. The same report found that only 26% of KEV-listed vulnerabilities were fully resolved, with a median remediation time of 43 days (up from 32 the year before).Organizations see the gap themselves. In the Zscaler ThreatLabz 2026 VPN Risk Report, 54% of organizations said they needed a week or more to patch a critical VPN vulnerability.Forty-three days, or even a week is a relatively long time when exploitation begins in a matter of days (if not sooner). This gap persists even in well-run security programs because the process, and decisions justifying that process is critical:Appliances must be scheduled for maintenance.Remote access can’t go down during business hours.High-availability pairs have to be upgraded in sequence.Fixes have to be tested against a configuration that has developed years of exceptions.Government guidance on patching vulnerabilities also reflects this operational reality. The Five Eyes intelligence partners issued a joint series on edge-device security in February 2025. CISA’s Binding Operational Directive 26-02 now orders federal agencies to replace edge devices that have reached end of support. In 2024, CISA, the FBI, and international partners jointly urged organizations to move away from traditional VPNs toward zero trust and secure access service edge (SASE) architectures. The Identity Stack Can’t Close the GapOver the past decade, most enterprises have invested heavily in identity-based solutions like single sign-on, multi-factor authentication (MFA), conditional access, and device posture checks. Those investments help organizations keep unauthorized users out of enterprise networks. In fact, Microsoft’s 2025 Digital Defense Report states that MFA can block over 99% of identity-based attacks.When we take a closer look at the failure modes attackers are utilizing, however, it’s clear that identity based solutions have a limited impact. Pre-authentication exploits never reach the identity provider. As far as the appliance is concerned, a forged or stolen session token is presented after authentication has already happened. Leaked credentials can be reused against a gateway that doesn’t enforce phishing resistant MFA. In each case the identity stack did its job, but the architecture offered the attacker a path that never touched it.This is the gatekeepers dilemma; a convenient way for legitimate users to access a network is also the most likely attack point, which is why convenience and ingress/egress features keep appearing in these security advisories. Persistent sessions, authentication override cookies, and single sign-on shortcuts exist because VPN reauthentication is painful for users. In a VPN architecture every convenience becomes a trust artifact stored or validated at the internet-facing edge. Administrators are forced to choose between a worse user experience and a larger more vulnerable attack surface.The consequences of that choice are visible in the ransomware data that Zscaler and other organizations have collected over the past decade. The Zscaler ThreatLabz 2026 Ransomware Report found that data exfiltration among the ten most active ransomware groups surged 275.8% year over year, to 896.2 terabytes. It’s Time to Change Where the Architecture SitsZero trust network access (ZTNA) doesn’t make software vulnerability-free, and no honest vendor should claim that it does. Every access platform, including its cloud brokers and endpoint clients, is software that needs secure engineering and rapid patching. But ZTNA changes the architecture those vulnerabilities exist in, and that changes everything.Where the trust decision lives: A VPN exposes a gateway and hands out network access. ZTNA exposes nothing and brokers access to one app at a time.Four principles separate these two architectures.Applications are invisible to the internet. In a cloud-native ZTNA model, private applications and the lightweight connectors in front of them make only outbound connections to a cloud exchange. Because nothing listens for inbound connections or exposes public IPs, applications are rendered completely invisible to the internet—leaving no portal for an attacker to discover, fingerprint, scan, or exploit.Access is segmented to specific applications, not the network. Rather than extending subnets or placing users on routed segments, a Zero Trust session connects authorized users, workloads, and AI agents directly to designated applications. By dynamically discovering private workloads and automating least-privilege user-to-app segmentation, lateral threat movement is eliminated. Even if a session is compromised, an attacker never receives an internal IP address or network routes.Trust is continuously verified and reinforced with active containment. Instead of relying on a forgeable bearer token or override cookie cached at an internet-facing gateway, identity, context, and device posture are evaluated continuously in the cloud throughout the session. Users get a seamless experience without perimeter shortcuts—and by integrating deception decoys directly into the access layer, organizations can immediately detect and isolate any post-compromise reconnaissance.Patching moves to the cloud. When the access layer is a multi-tenant cloud service, platform updates are deployed seamlessly without waiting for customer maintenance windows or complex HA upgrades. Modern Zero Trust architectures can also continuously scan private applications for specific exploit paths and apply precise inline virtual patches at machine speed, closing the 43-day remediation gap long before vulnerabilities can be weaponized.Here’s how each failure mode plays out under these two models:Failure modeTraditional VPNZero trust architecturePre-auth exploitationConcentrator listens for inbound connections on the public internetApplications and connectors make outbound-only connectionsSession bypassSession state is held and validated at the exposed edgeSessions are brokered per user and per application in the cloud, with continuous identity and posture checksCredential exposureCredentials and session secrets are handled on the applianceAuthentication is delegated to the identity provider; the access layer never stores user credentialsPersistence after patchingCustomer-operated appliance; attacker can hide in firmwareCloud-operated service with no customer-patched gateway to embed malwareNetwork-level accessSuccessful entry places the attacker on the networkUsers only connect to specific applications  Solving the Gatekeepers DilemmaCritical vulnerabilities like PAN-OS (CVE-2026-0257), FortiOS (CVE-2024-21762), PHP2Shell (CVE-2026-109618) and others are now endemic, and this trend will likely accelerate due to frontier AI-powered attacks. In addition to data loss and system compromise, CVEs take an enormous toll on organizations in the form of monitoring and patching complex global systems.Statistically, the next critical vulnerability is just a few weeks away. When it arrives, the response playbook will be the all-too-familiar emergency patching, certificate and credential rotation, log reviews, and a lingering concern that the attacker may have left something behind.Organizations can keep running that playbook indefinitely, or they can recognize that the playbook exists because the architecture demands it, and change the architecture.Administrators who enabled the features exploited in recent advisories weren’t being careless, they were trying to solve real user problems within the limits of a legacy architecture. But when dozens of VPN and firewall vulnerabilities have landed on CISA’s KEV catalog in just the last three years, it’s clear that organizations can’t patch their way out of an architectural design flaw. Zero trust solves the gatekeeper’s dilemma by removing exposed gateways altogether, to give users seamless and direct access to applications without exposing your corporate network to the internet.Zscaler Private Access (ZPA), the industry’s first and only AI-powered ZTNA solution, eliminates the gatekeeper’s dilemma by connecting authorized users directly to specific private applications through the Zscaler Zero Trust Exchange. Instead of defending an exposed perimeter gateway that leads to your corporate network, ZPA secures private applications across four core pillars:Hide: Applications are not exposed to the internet. Lightweight App Connectors establish inside-out connections, eliminating public IPs and exposed DMZ gateways so threat actors have nothing to scan, fingerprint, or attack.Segment: Access is brokered on a least-privileged, per-user and per-application basis, with continuous identity and device posture checks. AI-powered app segmentation dynamically recommends and enforces granular user-to-app policies that eliminate lateral movement.Shield: Real-time inline threat and data protection blocks exploits in transit, protecting and patching vulnerable private applications before attackers can strike.Deceive: Native deception and decoys lure, detect, and isolate sophisticated threat actors and malicious insiders attempting post-compromise reconnaissance.Because ZPA is delivered as a cloud-native service across 200+ locations, there is no VPN waiting for an emergency maintenance window when the next critical CVE surfaces.Ready to leave emergency VPN patching behind?Schedule a demo to see Zscaler in actionRead the ThreatLabz 2026 VPN Risk Report