Vulnerabilities

A WAF rule is not a patch: ShinyHunters returns to Oracle PeopleSoft flaw CVE-2026-35273

ShinyHunters resumed exploiting PeopleSoft CVE-2026-35273 by slipping past WAF rules kept in place of Oracle's June patch. Why it worked and what to fix.

All articles

On 25 September 2026, Mandiant and Google Threat Intelligence Group (GTIG) reported that the extortion group ShinyHunters, which Google tracks as UNC6240, had resumed mass exploitation of CVE-2026-35273, a critical Oracle PeopleSoft flaw that Oracle patched on 10 June. No new vulnerability was needed. The group percent-encoded a single character in the request path, so firewall rules written to block the vulnerable component no longer matched, while the PeopleSoft application server still decoded the request and passed it to the vulnerable code. Google says web shells were deployed on dozens of systems worldwide, in sectors from higher education and healthcare to transportation and government.

In the same weeks, the FBI was investigating ShinyHunters’ claim that it had broken into the bureau’s jobs portal through PeopleSoft. On 5 October the FBI’s cyber division told Reuters that the incident resulted from a platform managed by a third party, where a contractor “failed to implement a security patch explicitly issued to secure the platform.” The FBI has not named the vulnerability, and Google’s report does not discuss the FBI case. Together, though, the two stories make one point: a control placed in front of unpatched code buys time, and that time runs out when the attacker finds the gap.

What happened

  • 27 May to 9 June: UNC6240 exploits CVE-2026-35273 as a zero-day, mostly against higher-education institutions, according to Mandiant.
  • 10 June: Oracle publishes an out-of-band Security Alert. The flaw is in the Environment Management Hub (PSEMHUB) of PeopleSoft Enterprise PeopleTools 8.61 and 8.62, is exploitable over HTTP without authentication, scores 9.8 on CVSS 3.1 and can lead to remote code execution. Earlier, unsupported releases are likely affected too.
  • 12 June: CISA adds the flaw to its Known Exploited Vulnerabilities (KEV) catalog, with a remediation deadline of 15 June for US federal civilian agencies.
  • 22 September: ShinyHunters tells BleepingComputer it used a “new” PeopleSoft zero-day to reach FBI systems and move into FBI-managed AWS GovCloud, claiming 2 to 3 TB of data. BleepingComputer could not verify this; the FBI said only that it was investigating claims of unauthorized activity affecting FBIjobs.gov.
  • 25 September: Mandiant and GTIG publish their analysis of the renewed campaign.
  • 26 September: BleepingComputer reports that ShinyHunters admitted using the WAF bypass against FBI Jobs, while still claiming it also used a “NEW unknown vulnerability in the same PSEMHUB component.”
  • 5 October: Brett Leatherman, assistant director of the FBI’s cyber division, gives Reuters the contractor explanation and says the contractor has been removed. Reuters’ sources identified the platform as PeopleSoft and said the breach exposed sensitive personal details of thousands of bureau employees.

How the WAF rules were bypassed

In June, Mandiant advised organizations that could not patch or disable the hub at once to block external access to the PSEMHUB path at the perimeter, warning that WAF body-inspection rules alone were insufficient. Many WAF and reverse-proxy rules compare the literal request path before URL decoding. The attackers requested the same path with its first letter percent-encoded; the rules saw an unfamiliar string, and WebLogic decoded it and served the vulnerable servlet as usual. “The current campaign demonstrates that UNC6240 adapted to published defensive guidance, targeting organizations that implemented WAF rules but did not patch the vulnerability,” Mandiant wrote.

Targets typically first received five to 15 probing requests that returned operating-system details without writing any file. Then came either JSP web shells, with the request repeated so that every node behind a load balancer likely received a copy, or fileless command execution whose trace on the host is a shell process started by the WebLogic Java process. On Windows, a trojanized installer posing as a signed Light Alloy media-player setup loaded a credential-stealing backdoor Google calls SIDEEYE; on Linux, the attackers used the legitimate MeshAgent remote-management tool to keep access. A quarter of the observed commands ran as root or SYSTEM.

Why it matters

Virtual patches have a shelf life. A WAF signature encodes the defender’s guess about what the attack looks like. The WAF, the reverse proxy and the application server each normalize URLs in their own way, and every mismatch is a potential bypass; Google warns of other percent-encoded, mixed-case or non-normalized variants. Mandiant’s report came three and a half months after Oracle’s patch, time in which every organization relying on a WAF rule could have patched.

A late patch needs a compromise check. Exploitation began two weeks before the alert, and Google also describes intrusions in July. CISA’s KEV notice cites Binding Operational Directive 26-04, which also sets expectations for checking whether a system was compromised before it was patched. A server patched in September after months of exposure needs a hunt, not just an update.

Outsourcing the operation does not outsource the risk. In the FBI’s account, the fix existed; what failed was patch ownership on a platform someone else ran. Organizations that hand ERP and HR systems to integrators often receive a closed ticket, not evidence.

HR and finance systems are extortion targets. PeopleSoft holds HR, payroll and student records, the tables Mandiant tells defenders to audit for bulk queries. Mandiant says UNC6240 has “a well-established pattern of data theft extortion.”

For universities, enterprises and IT service providers in Vietnam that run PeopleSoft or similar ERP and HR suites, for themselves or for foreign clients, the question is not whether a WAF is in place. It is whether the patch is installed and verified on every node, and whether anyone has checked for compromise since late May.

When a WAF rule is an acceptable stopgap

A virtual patch is reasonable only while all four conditions hold:

  1. It has an owner and an expiry date measured in days for an actively exploited, unauthenticated, internet-facing flaw. CISA gave federal agencies three days for this one.
  2. It blocks the whole component on the normalized path, decoded and case-folded, instead of matching payloads, and it has been tested against encoded variants.
  3. The component does not need internet access in the first place. Mandiant notes that restricting the hub and the Integration Broker listening connector from the public internet does not break standard user sessions, so network-level blocking is better than a WAF rule.
  4. It comes with monitoring and a compromise assessment, because the rule does nothing about an intruder who arrived earlier.

Otherwise, the rule mainly produces a false sense of closure.

What to do now

  1. Patch and verify. Apply Oracle’s fix for CVE-2026-35273, stay on supported PeopleTools releases, and confirm the fix on every node by checking versions, not ticket status. Where the hub is not used, disable it or remove the PSEMHUB application, as Mandiant and Oracle advise.
  2. Take administrative components off the internet. Make the hub and the Integration Broker listening connector reachable only from management networks or a VPN, and inventory every internet-exposed ERP and HR endpoint.
  3. Fix the rules you keep. Match on normalized paths, test encoded and mixed-case variants, and log each remaining virtual patch as accepted risk with an expiry date.
  4. Hunt back to 27 May. Search WebLogic access logs for the PSEMHUB path and percent-encoded variants such as /%50SEMHUB/, especially external POST requests to the hub; check all nodes for unexpected .jsp, .jspx and .exe files in the PSEMHUB and PORTAL directories; alert on shell processes started by the WebLogic Java process; and look for unexpected MeshCentral agents.
  5. Treat a web shell as a full compromise. Preserve evidence, rotate every credential the PeopleSoft service account can read (database connection strings, Integration Broker credentials, cloud keys), and stop running WebLogic as root or SYSTEM.
  6. Watch for data leaving. Look for large archive files in temporary or web directories, compression or copy tools started by service accounts, bulk queries against HR, payroll and student tables, and sustained outbound transfers from the PeopleSoft tier.
  7. Write patch ownership into contracts. For outsourced ERP and HR platforms, set a patch deadline for actively exploited flaws, require evidence of installation, keep the right to verify, and require prompt incident notification.

Era Artech’s perspective

WAFs remain useful, and a well-built rule can cover the few days a patch cycle takes. The PeopleSoft campaign shows what happens when a temporary control is recorded as a permanent fix. We suggest recording two dates for every critical, exploited vulnerability (the day the patch was verified on all systems and the day the compromise check was completed) and treating the system as open risk until both exist, whoever operates the platform.

Sources

  1. Oracle Security Alert Advisory - CVE-2026-35273 Oracle ·
  2. CISA Adds One Known Exploited Vulnerability to Catalog Cybersecurity and Infrastructure Security Agency (CISA) ·
  3. ShinyHunters Renewed Mass Exploitation Campaign Targeting Oracle PeopleSoft Google Cloud (Mandiant and Google Threat Intelligence Group) ·
  4. ShinyHunters uses WAF bypass trick in Oracle PeopleSoft attacks BleepingComputer ·
  5. ShinyHunters claims FBI hack, data theft in PeopleSoft zero-day breach BleepingComputer ·
  6. Exclusive: Accenture contractor removed from FBI following damaging data breach, sources say Reuters ·

This analysis reflects information that was publicly available on its publication date. It is provided for general information only and is not legal or professional advice.

Need help assessing your exposure?

Our team can review your systems against the issues discussed here and help you prioritize fixes.

Talk to our security team