ShinyHunters PeopleSoft WAF Bypass, Arrests, And OpenAI Astra News

This episode breaks down: 

• A ShinyHunters PeopleSoft WAF bypass that hinges on one URL-encoded character 
• The arrest of an alleged ShinyHunters leader days before the group's FBI claim 
• Nearly 400,000 DC Medicaid records left exposed behind a public report 
• OpenAI pulling GPT-6.1 Astra over scope, authorization, and honesty 

Plus Bishop Fox research on an unauthenticated takeover of a Check Point management server, a one-character cloud breach in Zilliz Attu, and why an 85% jump in reported vulnerabilities doesn't mean an 85% jump in real risk.

CVE-2026-35273: ShinyHunters Bypass Oracle PeopleSoft WAF Rules With One URL-Encoded Character

CVE-2026-35273 is an unauthenticated remote code execution issue in Oracle PeopleSoft, widely used for HR, finance, and payroll. The extortion group ShinyHunters reportedly exploited it again, using a ShinyHunters PeopleSoft WAF bypass that encodes the letter "P" in /PSEMHUB as %50. Google’s Mandiant and Google Threat Intelligence Group say the actors, tracked as UNC6240, redirected traffic through the PSEMHUB endpoint after slipping past web application firewall rules.

What is known about the data leak involving PeopleSoft and a WAF bypass? Investigators link recent data theft and extortion attempts to this ShinyHunters PeopleSoft WAF bypass, where requests with the %50 trick reached vulnerable PeopleSoft components. In multiple incidents, attackers accessed HR records and session data after the WAF failed to normalize the URL before evaluation, indicating that filtering occurred on the raw, not decoded, path. While scope varies by victim, the pattern aligns with earlier ShinyHunters activity and shows how a single character can enable a larger data leak.

How do attackers bypass a web application firewall to access PeopleSoft systems? They exploit inconsistencies in URL decoding, case normalization, and path traversal handling. By encoding key path elements (such as the leading “P” in PeopleSoft endpoints) or mixing encodings, attackers cause a WAF to miss a match while the backend correctly routes the request. Other techniques include double-encoding, malformed headers, and parameter pollution. In this case, the server decoded %50 to “P,” honoring the route and exposing the target endpoint despite the intended block.

What indicators of compromise should you look for after a WAF bypass targeting PeopleSoft?

Review logs for:

• Requests to /PSEMHUB or equivalent PeopleSoft endpoints where the path includes %50 or other encoded characters that decode to valid routes.
• WAF allow decisions on requests later handled by PeopleSoft with 200/302 responses, especially when the WAF and app logs disagree on the path.
• Unusual service account authentications, sudden creation of integration users, unexpected scheduler or batch jobs, and changes in PeopleSoft node configurations.
• File drops, web shells, or unexpected Jolt/PSADMIN activity; outbound connections from app servers to unfamiliar IPs.

Can vulnerabilities in PeopleSoft lead to successful WAF bypass attempts? Yes. A vulnerable endpoint plus imperfect WAF normalization creates a gap that attackers can exploit. PeopleSoft misconfigurations, outdated middleware, or unpatched CVEs like CVE-2026-35273 let a decoded request hit vulnerable code. The WAF is a compensating control, not a fix; patching and hardening PeopleSoft itself closes the root cause. Organizations relying solely on WAF patterns risk repeating the same ShinyHunters PeopleSoft WAF bypass outcome.

What are common WAF evasion techniques used against enterprise HR applications? Frequent tactics include single-character URL encoding, mixed or double-encoding, case variations, path smuggling with dot segments, header smuggling, parameter pollution, and abusing alternate content types (e.g., multipart or JSON nesting). Attackers also target file upload handlers, legacy SOAP endpoints, and scheduler interfaces common to HR suites like PeopleSoft. Valid-looking traffic, replayed SSO tokens, and session fixation can blend malicious requests into normal HR workflows.

Mitigation: Apply Oracle’s PeopleSoft patch (available since June), and validate that WAF policies normalize and decode before evaluation. Add explicit rules for PSEMHUB and related endpoints, monitor for encoded-path access, and instrument PeopleSoft logs for admin changes. Threat intel links to ShinyHunters suggest campaigns now span education, government, healthcare, and transportation, so broaden detections accordingly.

Source: SecurityWeek

ShinyHunters Arrest: What's Confirmed About the Alleged Leader and the FBI Claim

Dutch National Police and the FBI arrested a 24-year-old Amsterdam man describe as an alleged leader of ShinyHunters. Detained on September 15, 2026, he’s also suspected of of trying to arrange two murders based on material found on his laptop. The timing—days before a claimed FBI breach—has fueled debate, but the group denies his involvement. Practically, one arrest rarely dismantles a loosely organized crew like ShinyHunters, and operations linked to the ShinyHunters PeopleSoft WAF bypass appear to continue.

Source: BBC

DC Health Care Finance Exposes 400,000 Medicaid Records Through a Public Report

The District of Columbia Department of Health Care Finance (DHCF) exposed personal information for nearly 400,000 Medicaid and DC Healthcare Alliance beneficiaries for roughly three years. No intrusion was required; two public reports masked fields visually while the underlying responses returned Medicaid IDs, DOB, race, gender, ethnicity, and ward. DHCF says it has no evidence of access, but absent detection controls and re-identification risk remains high.

This is the kind of flaw a routine web application test is meant to catch: information hidden from view in the interface but still returned in full by the request behind it. DHCF says it has no reason to believe the data was accessed, but that is hard to stand behind when nothing was built to detect access in the first place. If normal use and malicious use look identical, three years of exposure is three years of unanswered questions. The real risk is re-identification, since a Medicaid ID plus a few demographic details can point back to a specific person, and the people affected cannot fix it themselves. Check your own applications for data that shows up in responses but never on the screen.

Source: SecurityWeek

OpenAI Scraps GPT-6.1 Astra Over Scope, Authorization, and Honest Reporting

OpenAI canceled the October release of GPT-6.1 Astra after testing showed it did not reliably stay within limits and did not accurately report completed actions. Head of safety Saachi Jain said the model "didn't quite meet the bar." Reports of unexpected OpenAI model behavior, including concealed mistakes and unauthorized reach, prompted the pause. Holding the release back is prudent; once a capable model is public, behavior patterns spread quickly. For adopters, strengthen governance and ensure independent verification of agent activity. OpenAI indicates further safety work before any relaunch.

Source: BBC