Security Bulletin

CVE-2026-35273: ShinyHunters Resumes Oracle PeopleSoft Exploitation with WAF Bypass

7 min read Updated

ShinyHunters(UNC6240) has resumed mass exploitation of CVE-2026-35273, a critical unauthenticated remote code execution vulnerability in Oracle PeopleSoft. The new campaign bypasses WAF rules and targets organizations that did not patch.

CVE-2026-35273 is a critical vulnerability in PeopleSoft Enterprise People Tools. Unauthenticated attackers can exploit it remotely to achieve remote code execution. The same group first exploited it as a zero-day between May 27 and June 9, 2026, mainly against universities and colleges. Those attacks led to data theft, persistence, and extortion. Oracle released an out-of-band patch on June 10.

In September 2026, the group returned. Attackers modified the exploit to bypass WAF rules that block the vulnerable endpoint. The new wave targets organizations that relied on those rules and did not patch. Victims now span technology, healthcare, government, and other sectors. Read on to understand how the exploit and WAF bypass work, and how AppTrana protects against both. 

Risk Analysis

What is CVE-2026-35273?

Severity
Critical
CVSS v3.1 Score
9.8
Exploitation Complexity
Low
Exploit Publicly Available
Yes
Attack Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

CVE-2026-35273 is a critical vulnerability affecting Oracle PeopleSoft Enterprise PeopleTools versions 8.61.x and 8.62.x. Oracle assigned it a CVSS score of 9.8. Attackers can exploit it remotely without authentication or user interaction. 

The vulnerability has been classified as a Server-Side Request Forgery (SSRF) vulnerability in the Environment Management component. Mandiant’s latest analysis shows attackers achieving code execution through Java deserialization in the PSEMHUB hub servlet. A separate pre-authentication vulnerability, CVE-2026-35278, was also disclosed in PeopleSoft People Tools. Successful exploitation leads to Remote Code Execution (RCE) and full control of the underlying server. 

Oracle released an out-of-band Security Alert on June 10, 2026. The July 2026 Critical Patch Update delivers the permanent fix for both CVE-2026-35273 and CVE-2026-35278. Any PeopleTools instance still unpatched should be treated as potentially compromised. 

The campaign shows how attackers adapt to published defensive guidance. Organizations that deployed WAF rules in place of patches became the next target within months. 

Latest activity

September 2026 Update: WAF Bypass Enables Renewed Exploitation 

In June, Mandiant advised organizations that could not patch immediately to block external access to /PSEMHUB/* at the perimeter. Many teams did this with path-based WAF or reverse proxy rules. UNC6240 then modified its exploit to evade these rules and target the servers that were still unpatched.

The attackers now request /%50SEMHUB/hub in place of /PSEMHUB/hub. %50 is the URL-encoded form of the letter P. A rule that matches the raw request path finds no match and passes the request. WebLogic decodes the path and routes the request to the vulnerable servlet. 

This is a URL normalization mismatch. The security control and the application read the same request differently. Any rule that inspects the path before decoding is exposed. Attackers can also use mixed-case or other non-normalized forms of the path. 

The campaign targets organizations that deployed WAF rules and did not patch. Web shells have been confirmed on dozens of systems across higher education, technology, IT services, healthcare, agriculture, transportation, and government.

Technical Analysis

How CVE-2026-35273 Works: SSRF to RCE Attack Chain

Security researchers have described CVE-2026-35273 as an SSRF-to-RCE vulnerability chain.

Server-Side Request Forgery vulnerabilities occur when attackers can manipulate a server into making requests on their behalf. In many cases, SSRF vulnerabilities are used to access internal services, cloud metadata endpoints, or otherwise inaccessible resources.

In the case of CVE-2026-35273, the SSRF capability appears to act as the initial stage of a broader exploitation chain that ultimately enables remote code execution.

Researchers identified two PeopleSoft endpoints that play a central role in exploitation:

  • /PSEMHUB/hub
  • /PSIGW/HttpListeningConnector

These components are associated with PeopleSoft’s Environment Management infrastructure and Integration Gateway functionality. Attackers leveraged these exposed interfaces to execute the attack chain and gain unauthorized access to targeted environments.

The involvement of these management components is significant because administrative and integration services often operate with elevated privileges and broad access across enterprise environments.

In the September campaign, attackers reached the same servlet through an encoded path, /%50SEMHUB/hub. Security teams should treat any encoded, mixed-case, and other non–normalized varients of the /PSEMHUB/hub path as equivalent. 

Attack Lifecycle

How the September Campaign Unfolds 

Mandiant observed a consistent sequence across targeted PeopleSoft environments. 

Stage 01: Target Verification

Attackers send 5 to 15 POST requests to /%50SEMHUB/hub containing a serialized Java object. Unpatched servers respond with the host operating system. No files are written and the service is not disrupted. This lets attackers quietly confirm a server is exploitable. Some organizations may see these requests in logs with no follow-on activity. 

Stage 02: Exploitation

Attackers use two methods. Both abuse Java deserialization in the hub servlet. 

  • Web shell deployment. Attackers send a burst of POST requests, then create new JSP files in the PSEMHUB.war directory. The repetition likely places a web shell on every node behind a load balancer. 
  • Fileless command execution. Command output returns directly in the HTTP response. No file is written to disk. On the host, the WebLogic Java process spawns cmd.exe or /bin/sh. 
Stage 03: Web Shells

Attackers deploy two single-line JSP web shells built to evade detection: 

  • x.jsp runs commands on Windows and Linux. It accepts hex-encoded commands over POST and hides the Linux shell path from static signatures. 
  • u.jsp (and a variant, u2.jsp) uploads files in 150 KB Base64 chunks. This bypasses request-size limits. It can also run commands once the upload is complete. 
Stage 04: SIDEEYE Backdoor

On Windows servers, attackers upload and run Ple64.exe. The file poses as a signed installer for the Light Alloy media player. It carries a valid Extended Validation certificate. The installer runs a three-stage chain that loads the SIDEEYE backdoor in memory. 

SIDEEYE communicates with its command-and-control server over raw TCP on ports 3333 and 3334. It supports browser and application credential theft, process and file management, and interactive reverse shell and reverse proxy access. 

Credential Exposure

SMB-Based Credential Exposure Risks

Researchers observed that exploitation attempts may trigger outbound SMB connections from vulnerable servers. When a vulnerable server initiates outbound SMB connections, attackers may capture Windows machine-account NetNTLM hashes. 

Captured authentication material can be used in relay attacks, credential abuse, or lateral movement. 

Organizations that permit unrestricted outbound SMB traffic from application servers face increased risk. They should review network controls as part of their mitigation strategy. 

Threat Intelligence

Post-Exploitation Activity Following Initial Access

Activity 01

Persistence via Remote Management Agents

Attackers deployed Mesh Agent, a remote management agent associated with the Mesh Central platform, to maintain access to affected systems. Some were disguised as Microsoft Azure-related services to avoid detection. In the September campaign, they also used JSP web shells, the SIDEEYE backdoor, and Neo-ReGeorg for tunneling. 

Activity 02

Internal Reconnaissance and Lateral Movement

Attackers searched for additional systems and accounts within the environment. Evidence of lateral movement shows attempts to extend access beyond the compromised PeopleSoft server. 

Activity 03

Data Exfiltration

Attackers compressed stolen data before transferring it out of victim environments. This indicates a structured effort to collect and remove information after gaining access. 

Detection

How to Detect CVE-2026-35273 Exploitation

Detection requires reviewing network, host, and application logs for signs of unauthorized activity. 

Review WebLogic access logs for POST requests to /PSEMHUB/hub, /PSIGW/HttpListeningConnector, and any percent-encoded or mixed-case variant such as /%50SEMHUB/hub. Treat a burst of 5 to 15 POST requests to the hub servlet as a likely exploitability check. Requests containing localhost, loopback, or internal IP references may indicate SSRF abuse. 

On affected systems, look for unexpected files in the PSEMHUB.war directory, including x.jsp, u.jsp, u2.jsp, tunnel.jsp, tunnel.jspx, and Ple64.exe. Watch for the WebLogic Java process spawning cmd.exe or /bin/sh. This catches fileless execution, which leaves no JSP file behind. Check every WebLogic node behind a load balancer. 

Review outbound traffic from PeopleSoft servers. Look for SMB connections to external systems and connections to TCP ports 3333 and 3334. Mandiant identified 162[.]219[.]30[.]165 as a SIDEEYE command-and-control server. Look for unexpected MeshCentral or MeshAgent processes. 

Review historical logs across both attack windows: May 27 to June 9 and September 2026 onward. Organizations that relied on WAF rules without patching should assume exposure and investigate. 

Mitigation

How to Mitigate CVE-2026-35273

Organizations running PeopleTools 8.61 or 8.62 should patch immediately. Confirmed exploitation across two campaigns makes any delay high risk. 

Step 01

Apply Oracle’s Security Update

Apply Oracle’s June 10 Security Alert patch and the July 2026 Critical Patch Update. WAF rules and path-based blocking are not a substitute for patching. 

Step 02

Review and Disable EMHub Where Not Required

Review whether the Environment Management Hub (EMHub) service is required. Oracle recommends disabling the service in multi-server deployments. In single-server environments, remove the PSEMHUB application where it is not needed. 

Step 03

Restrict Access to Vulnerable Endpoints

Block external access to /PSEMHUB/* and /PSIGW/HttpListeningConnector  as a temporary control. Rules must match the decoded, normalized path. Rules that match the literal string can be bypassed with encoding. Monitor outbound SMB traffic and connections to TCP ports 3333 and 3334 from PeopleSoft servers. 

Step 04

Conduct a Post-Patch Compromise Assessment

Exploitation occurred before and after Oracle’s patch. Conduct a compromise assessment after remediation. Review web application directories, recently modified files, unexpected JSP files, unauthorized directories, and outbound network connections. Inspect all WebLogic nodes, since attackers repeated requests to plant web shells on every node behind a load balancer.

Step 05

Rotate Exposed Credentials 

Rotate credentials readable by the PeopleSoft service account. This includes database connection strings in psappsrv.cfg, Integration Broker credentials, and cloud credentials reachable from the web tier. 

AppTrana Protection

How AppTrana Reduces Exposure to CVE-2026-35273

AppTrana helps protect against exploitation attempts associated with CVE-2026-35273 by detecting and blocking malicious requests targeting vulnerable PeopleSoft endpoints, including SSRF and remote code execution attack patterns.  

It also decodes and normalizes request paths before rule evaluation. Requests to /%50SEMHUB/hub and other encoded forms of the PSEMHUB path are evaluated as /PSEMHUB/hub and blocked.   

AppTrana reduces exposure while teams apply Oracle’s patches. Autonomous protection, with expert-backed operations built in, blocks exploit attempts at the edge while you patch and complete your compromise assessment. 

 The screenshots below show AppTrana successfully detecting and blocking CVE-2026-35273 exploitation attempts, returning an HTTP 406 (Not Acceptable) response: 

AppTrana blocking exploit attempt on /psemhub/hub endpoint, CVE-2026-35273
AppTrana blocking exploit attempt on /PSIGW/HttpListeningConnector endpoint, CVE-2026-35273

AppTrana blocking encoded /%50SEMHUB/hub bypass attempt, CVE-2026-35273

AppTrana blocking encoded /%50SEMHUB/hub bypass attempt, CVE-2026-35273

Stay tuned for more relevant and interesting security articles. Follow Indusface on Facebook, Twitter, and LinkedIn.
Nikita Waghole
Nikita Waghole

Nikita is a Cyber Security Researcher specializing in Web Application Firewall (WAF) technologies and web threat detection. She has hands-on experience in analysing traffic, identifying attack patterns, and developing security rules. she plays a key role in strengthening application security and mitigating real-world attacks. She focuses on vulnerability analysis, false positive tuning, and strengthening application security against real-world threats. She actively investigates emerging threats, including zero-day vulnerabilities and automated attack patterns, ensuring robust protection for modern web applications. Passionate about cybersecurity, Nikita continuously enhances her skills to build more effective and reliable defence mechanisms.