Weekly Report: New Hacking Techniques and Critical CVEs 07 Sep to 13 Sep 2026

CISA added 14 vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog across 11 vendors between 07 and 13 September 2026, 8 of them rated CVSS 9.0 or above, and public proof-of-concept code is already published for at least 5, including two flaws disclosed and exploited in the same week. Patch Adobe Commerce and Magento first: CVE-2026-75650 is already being used to plant backdoors on live stores, not just theorized. Three of the five vulnerabilities below sit on internet-facing management or DevOps infrastructure, which is the pattern to watch: attackers are going after the systems that manage other systems, not just the applications end users touch.

Which vulnerabilities should you patch first this week?

• CVE-2026-75650, Adobe Commerce and Magento template engine RCE. This ranks first because exploitation is not theoretical, it is already producing victims: attackers have been deploying a Rust-based Linux backdoor and a PHP web shell on compromised stores since 4 September 2026, with one observed compromise completed within 50 minutes of the first recorded attack. Apply Adobe’s hotfix VULN-39341 immediately.
• CVE-2026-86218, N-able N-central pre-authentication RCE. This ranks second because N-central is remote monitoring and management software that MSPs use to reach client networks, so one compromised N-central instance is a lateral-movement path into every environment it manages. Huntress detected a customer compromise on 4 September 2026, and watchTowr independently reproduced the exploit. Apply hotfix 2026.3.1.14 now.
• CVE-2026-20079, Cisco Secure Firewall Management Center authentication bypass. This ranks third because it grants root-level control of the underlying operating system with no authentication, on a device whose entire purpose is managing an organization’s firewall fleet, and Cisco Talos has now tied two separate intrusion clusters to it, one with tooling overlapping the Sandworm APT and one matching Qilin ransomware tactics. Apply Cisco’s hotfix before the 12 September 2026 federal deadline passes further behind.

CVE-2026-75650: Adobe Commerce and Magento template engine remote code execution

What it is: An unauthenticated remote code execution flaw in Adobe Commerce and Magento’s template engine, caused by improper neutralization of special elements in template processing, allowing an attacker to trigger arbitrary code execution with no authentication and no user interaction.

Affected products and versions: Adobe Commerce 2.4.4 through 2.4.9, Adobe Commerce B2B 1.3.3 through 1.5.3, and Magento Open Source 2.4.6 through 2.4.9.

Exploitation status: Confirmed in the wild. Active exploitation began 4 September 2026 at 10:20 p.m. UTC, with attackers using the flaw to generate a “Payment Transaction Failed Reminder” email that triggers PHP code injection. CISA added CVE-2026-75650 to KEV on 8 September 2026. Federal remediation due date: 11 September 2026.

Patch status: Adobe released hotfix VULN-39341 on 8 September 2026, available through repo.magento.com/patch/VULN-39341-composer-patches.zip.

What to do: Apply VULN-39341 immediately regardless of whether your store shows signs of compromise. Search for the Rust-based Linux backdoor and PHP web shell attackers have deployed post-exploitation, since a patched server can still be hosting an earlier implant. Disable or restrict the transactional email templates the exploit abuses until the patch is confirmed applied fleet-wide.

Why it matters: An unauthenticated attacker who reaches this flaw gets full code execution on the server running your storefront, and the observed 50-minute time-to-compromise means the gap between disclosure and exploitation has effectively closed for widely deployed ecommerce platforms. Every unpatched Commerce or Magento instance reachable from the internet should be treated as compromised until proven otherwise, not merely at risk.

CVE-2026-86218: N-able N-central pre-authentication remote code execution

What it is: A static code injection vulnerability in N-able N-central, a remote monitoring and management (RMM) platform used by managed service providers, allowing unauthenticated remote code execution against the N-central server itself.

Affected products and versions: N-able N-central, all versions before 2026.3.1.14.

Exploitation status: Confirmed in the wild. Huntress detected a customer compromise on 4 September 2026, and watchTowr independently reproduced the exploit end to end. CISA added CVE-2026-86218 to KEV on 8 September 2026. Federal remediation due date: 11 September 2026.

Patch status: Fixed in N-central 2026.3.1.14, released as an emergency hotfix on 5 September 2026.

What to do: Apply 2026.3.1.14 immediately and treat any N-central server that was internet-reachable before the patch as a suspected pivot point, not just a patched asset. Rotate credentials that N-central holds for downstream managed endpoints. Review N-central’s outbound and lateral connections to client networks for anomalous activity dating back to at least 4 September 2026.

Why it matters: RMM software exists specifically to hold privileged access into many client networks at once, so a pre-auth RCE here is not a single-tenant incident, it is a template for reaching every downstream network the platform manages. This is the same class of third-party platform risk that has driven the largest MSP-linked ransomware campaigns of the past several years.

CVE-2026-20079: Cisco Secure Firewall Management Center authentication bypass

What it is: An authentication bypass in Cisco Secure Firewall Management Center’s web interface, caused by an improperly created system process at boot, that lets an unauthenticated attacker execute script files and obtain root access to the underlying operating system.

Affected products and versions: Cisco Secure Firewall Management Center Software, versions 7.0.0 through 10.0.1, spanning more than 70 point releases across the 7.0.x through 7.7.0 and later branches.

Exploitation status: Confirmed in the wild. Cisco Talos identified three separate intrusion clusters exploiting this flaw, including one, tracked as UAT-11823, with tooling overlapping the Sandworm APT group with high confidence, and another, UAT-11988, using tactics consistent with Qilin ransomware affiliates. CISA added CVE-2026-20079 to KEV on 9 September 2026. Federal remediation due date: 12 September 2026.

Patch status: Cisco has released hotfixes for affected versions; a comprehensive hardening release is scheduled for the week of 16 September 2026.

What to do: Apply the current hotfix now rather than waiting for the hardening release. Remove FMC’s web management interface from direct internet exposure wherever that exposure is not operationally required. Hunt for Talos’s published indicators of compromise for both the Sandworm-linked and Qilin-linked clusters.

Why it matters: FMC manages an organization’s firewall fleet, so root access here can extend an attacker’s reach into every firewall the console administers, not just the management server. The same flaw being used for both suspected espionage and ransomware in the same week is a concrete illustration of why continuous validation of internet-facing management infrastructure matters more than an annual check on a device most teams assume is already secured.

CVE-2026-85706: GitLab Community and Enterprise Edition unauthenticated path traversal

What it is: A path traversal vulnerability in GitLab’s repository commits API, caused by improper path confinement and missing authentication enforcement, that lets an unauthenticated attacker read arbitrary files from a GitLab server with no user interaction.

Affected products and versions: GitLab Community Edition and Enterprise Edition, versions 18.7 up to but not including 19.1.8, 19.2 up to but not including 19.2.6, and 19.3 up to but not including 19.3.2.

Exploitation status: Confirmed in the wild, with reporting indicating exploitation began within a day of disclosure. CISA added CVE-2026-85706 to its Known Exploited Vulnerabilities catalog on 11 September 2026. Federal remediation due date: 14 September 2026. Public proof-of-concept exploit code is published on GitHub.

Patch status: Fixed in GitLab 19.1.8, 19.2.6, and 19.3.2.

What to do: Patch to the fixed version immediately, since public PoC code lowers the skill required to exploit this to nearly zero. Rotate any secrets, tokens, or SSH keys that could have been readable through arbitrary file access on affected instances. Audit access logs for the commits API for requests to unusual file paths predating the patch.

Why it matters: An unauthenticated file-read on a self-hosted web application or API platform can expose configuration files, credentials, and CI/CD secrets that turn a read-only bug into full repository or pipeline compromise, which is why the practical severity here matches the CVSS 10.0 score rather than reading as inflated.

CVE-2026-19490: Citrix NetScaler ADC and Gateway authentication bypass

What it is: An authentication bypass in NetScaler ADC and NetScaler Gateway, exploited via the SAML HTTP-Redirect binding, that allows an attacker to bypass authentication without valid credentials.

Affected products and versions: NetScaler ADC and NetScaler Gateway, versions 13.1 through 13.1-63.21 and 14.1 through 14.1-73.32.

Exploitation status: Confirmed in the wild, with reporting that exploitation followed shortly after a public proof-of-concept was released. CISA added CVE-2026-19490 to KEV on 9 September 2026. Federal remediation due date: 12 September 2026.

Patch status: Citrix has released fixed builds for affected NetScaler ADC and Gateway versions.

What to do: Patch to the fixed build immediately rather than relying on the mitigations Citrix published for earlier NetScaler flaws, which do not apply to this authentication bypass. Terminate active sessions and force re-authentication after patching, since a bypass exploited before the patch could leave a valid session in place. Review NetScaler Gateway logs for authentication events that skipped expected SAML validation steps.

Why it matters: NetScaler Gateway is frequently the first externally reachable system in a remote-access chain, so an authentication bypass here hands an attacker the same starting position as a valid VPN user, without a password, credential stuffing attempt, or MFA prompt to trip alerts. Edge authentication systems like this one are exactly the class of asset that needs testing on the same cadence attackers use, not once a year.

New hacking technique: AI-orchestrated mass exploitation of PaperCut print servers

What researchers demonstrated: GreyNoise published research on 9 September 2026 showing that a single attacker used hundreds of autonomous AI agents, orchestrated with OpenAI’s Codex and executed using a DeepSeek model, to exploit PaperCut NG/MF print management servers across at least 395 organizations in a single campaign.

How it works:

1. AI agents synthesized publicly available information about two PaperCut vulnerabilities, CVE-2026-81578 (CVSS 9.8, authentication bypass) and CVE-2026-82078 (CVSS 9.1, unsafe reflection leading to code execution), both already known to CISA KEV before this campaign.
2. The DeepSeek model, chosen specifically for having fewer content-safety restrictions than mainstream alternatives, generated and refined working attack code.
3. Agents validated the exploit chain against the attacker’s own private lab environment before deployment.
4. Agents scanned the internet using Netlas.io to enumerate reachable PaperCut servers at scale.
5. Parallel agent deployment against live targets achieved first remote code execution in under four hours and domain administrator access two hours after that, using established techniques including LSASS credential harvesting, pass-the-hash, and the noPac attack against unpatched Active Directory environments.
6. Agents performed post-exploitation triage automatically, selecting attack paths based on each victim’s specific environment and running DCSync operations to exfiltrate domain credentials.

Why it matters: This is not a new vulnerability class, it is a new speed and scale profile applied to already-known vulnerabilities and already-known post-exploitation techniques. The four-hour time to initial access and six-hour time to domain admin compress what previously required a coordinated human team into a single operator’s afternoon, which changes the planning assumption security teams can make about how much warning time a newly disclosed CVE actually buys them.

What to do: Treat any internet-facing PaperCut NG/MF instance as a priority patch target regardless of whether it has shown signs of compromise, since these two CVEs were already in KEV before this campaign began. Disable or tightly restrict LSASS access and monitor for pass-the-hash and DCSync activity on domain controllers reachable from print management infrastructure. Assume that any publicly known, unpatched vulnerability now has a meaningfully shorter real-world exploitation window than in prior years, and plan patch cycles accordingly.

CISA KEV additions, 07 to 13 Sep 2026

No KEV additions were recorded for 7, 12, or 13 September 2026 based on available CISA alert pages for those dates; both the 12th and 13th fell on the weekend.

Related reading

• Weekly Report: New Hacking Techniques and Critical CVEs, 31 Aug to 6 Sep 2026: covered the original disclosure of the two PaperCut CVEs this week’s AI-orchestrated campaign exploited, along with the Kestra and SonicWall SMA1000 flaws from the prior window.
• The Great AI Divide: background on the widening gap between attacker speed and defender testing cadence that this week’s four-hour AI-orchestrated compromise illustrates directly.

How FireCompass helps

Every CVE in this week’s roundup sits on software that is supposed to be internet-facing and trusted by design: a firewall console, a source code platform, an ecommerce engine, an RMM tool, a VPN gateway. Knowing a CVE exists is not the same as knowing whether your specific, configured instance is actually reachable and exploitable, which is the gap between a vulnerability scanner’s output and a real finding. FireCompass’s agents beat top human researchers in internal evaluation 60 to 70 percent of the time at exactly this kind of exposure validation, turning a KEV list into a prioritized, evidence-backed patch order instead of a spreadsheet of CVE IDs.