Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-113604 Overview
CVE-2026-113604 is a denial of service vulnerability in Ghidra, the open-source software reverse engineering suite maintained by the National Security Agency. The flaw affects Ghidra versions 10.2 through 12.1.4 and resides in the ElfHeader class inside the ELF loader. The class allocates arrays sized by unvalidated extended section and program header counts read directly from the ELF file. A specially crafted 128-byte ELF file forces Ghidra to request arrays that exceed available Java heap space, triggering an OutOfMemoryError during load-spec discovery at import time or inside headless analysis pipelines.
Critical Impact
A single malformed ELF binary can crash Ghidra import jobs and halt batch headless analysis workflows used by malware triage teams.
Affected Products
- Ghidra 10.2 through 12.1.4 (ELF loader ElfHeader class)
- Ghidra headless analyzer pipelines that auto-import ELF samples
- Downstream automation that embeds Ghidra for ELF triage
Discovery Timeline
- 2026-10-20 - CVE-2026-113604 published to NVD
- 2026-10-20 - Last updated in NVD database
Technical Details for CVE-2026-113604
Vulnerability Analysis
The vulnerability is categorized as [CWE-789: Memory Allocation with Excessive Size Value]. During ELF parsing, Ghidra reads the extended section header count and extended program header count fields and uses them directly to allocate backing arrays. The ELF loader does not validate these counts against the actual file size before allocation. A 128-byte file is large enough to populate the ELF header fields that drive array sizing, yet far too small to contain the headers those counts imply. When Ghidra attempts to allocate the resulting multi-gigabyte arrays, the Java Virtual Machine raises OutOfMemoryError, aborting the import or headless analysis task.
Root Cause
The root cause lives in ElfHeader.java within ghidra.app.util.bin.format.elf. The parser trusts the extended header count fields without cross-checking them against the readable byte range provided by the BinaryReader. There is no upper bound sanity check and no EOFException guard before the allocation request reaches the JVM.
Attack Vector
Exploitation requires an analyst or automated pipeline to import a crafted ELF file. The attack vector is local and requires user interaction, consistent with analyst-driven reverse engineering workflows. Attackers can seed malware repositories, sandbox feeds, or shared sample sets with the malformed ELF to disrupt triage. Automated headless Ghidra pipelines that ingest untrusted samples are particularly exposed because the crash can halt an entire batch job.
// Patch excerpt from ElfRelocation.java hardening
package ghidra.app.util.bin.format.elf;
import java.io.EOFException;
import java.io.IOException;
import java.lang.reflect.Constructor;
import ghidra.app.util.bin.BinaryReader;
import ghidra.app.util.bin.StructConverter;
import ghidra.app.util.bin.format.elf.extend.ElfLoadAdapter;
import ghidra.app.util.bin.format.elf.relocation.AbstractElfRelocationHandler;
import ghidra.app.util.bin.format.elf.relocation.ElfRelocationContext;
import ghidra.program.model.data.*;
Source: Ghidra Commit 6c50a66 — the commit introduces EOFException handling and tightens parse validation for program and section headers.
Detection Methods for CVE-2026-113604
Indicators of Compromise
- ELF files with a total size under 1KB that declare non-zero extended section or program header counts in the ELF header.
- Ghidra import logs containing java.lang.OutOfMemoryError originating in ghidra.app.util.bin.format.elf.ElfHeader.
- Repeated headless analyzer process terminations correlated with specific ELF samples pulled from untrusted sources.
Detection Strategies
- Pre-screen ELF submissions with a lightweight parser that validates e_shnum, e_phnum, and the SHT_NULL extended counts against actual file length.
- Monitor JVM telemetry on analyst workstations and headless worker nodes for sudden heap exhaustion tied to Ghidra processes.
- Hash and quarantine samples that caused prior Ghidra crashes to prevent re-ingestion by batch pipelines.
Monitoring Recommendations
- Alert on Ghidra process exit codes paired with OutOfMemoryError entries in the application log.
- Track sample provenance for ELF files imported into shared Ghidra project repositories to identify the submission source after a crash.
- Baseline normal Ghidra memory usage so anomalous spikes during ELF load-spec discovery are flagged quickly.
How to Mitigate CVE-2026-113604
Immediate Actions Required
- Upgrade Ghidra to the fixed release that includes commit 6c50a66e8a939e2bcbaf0a6432ec5d53a41ceeb3 (GP-7250 parse hardening).
- Isolate headless analysis workers so a crash on one sample does not stall queued jobs for other analysts.
- Validate all ELF samples pulled from untrusted feeds with a stand-alone pre-parser before Ghidra ingestion.
Patch Information
The fix is tracked as GP-7250 Improved ElfHeader parse hardening for program headers and section headers. See the Ghidra Commit Log, the updated Ghidra ELF Header Code, and the VulnCheck Advisory on Ghidra for full remediation guidance.
Workarounds
- Reject ELF files smaller than the minimum size required to hold the declared number of program and section headers.
- Run Ghidra imports inside containers with capped memory and automatic restart so a single malicious sample cannot stall batch pipelines.
- Restrict the ability to add ELF samples to shared Ghidra projects to vetted users while patching is in progress.
# Pre-import sanity check for ELF header counts
filesize=$(stat -c%s "$1")
if [ "$filesize" -lt 1024 ]; then
echo "[!] Suspicious ELF under 1KB: $1" >&2
exit 1
fi
readelf -h "$1" | grep -E "Number of (program|section) headers"
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

