A critical heap out-of-bounds read and write vulnerability in Tesseract OCR versions 5.5.3 and earlier poses a significant risk of memory corruption and arbitrary code execution.

The vulnerability exists within the FullyConnected::DeSerialize function in src/lstm/fullyconnected.cpp, where insufficient validation of deserialized layer scalars leads to out-of-bounds memory access. An unauthenticated attacker can trigger this flaw by providing a maliciously crafted .traineddata file, leading to heap corruption or potential controlled execution.

Successful exploitation of this vulnerability could lead to a complete compromise of the process running the OCR engine. Given the high CVSS score of 8.6, the risk of heap corruption, system crashes, or potential unauthorized data access is severe, particularly for enterprise applications that process untrusted documents automatically.

Immediate Action: As no official patch is currently available, organizations should isolate systems that process external or untrusted .traineddata files. Monitor the Tesseract GitHub security advisory page for the release of a patched version.

Proactive Monitoring: Implement robust input validation for any user-supplied .traineddata files before they are processed by the Tesseract engine. Review system logs for unexpected crashes or segmentation faults associated with the Tesseract process.

Compensating Controls: Deploy sandboxing or containerization with restricted memory access to limit the impact of a potential heap corruption event. Ensure that the Tesseract process runs with the least privilege necessary to perform its functions.

Public Exploit Available: Unknown

Due to the severity of this heap corruption vulnerability, organizations must treat this as a high-priority risk. Until a formal security update is provided by the Tesseract project, restrict the ability of the OCR engine to ingest files from untrusted sources to prevent potential exploitation.