These pages are published after PatchSiren validates generated defensive summaries against stored public CVE and source evidence.
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-09-10T18:18:14.600Z and has not been modified since then. The NVD entry is currently Analyzed. Tesseract OCR engine versions 5.5.3 and earlier are vulnerable to a denial of service attack due to a flaw in the Plumbing::DeSerialize function in src/lstm/plumbing.cpp. This function accepts a zero-length s [truncated]
CVE-2026-88053 is a high-severity vulnerability in Tesseract, an open-source OCR engine. The vulnerability exists in version 5.5.3 and earlier, where the Classify::ReadIntTemplates function in src/classify/intproto.cpp reads values from a crafted .traineddata file without validation, leading to heap out-of-bounds pointer writes, heap corruption, a crash, or potentially controlled corruption.
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-09-10T18:18:14.317Z and has not been modified since then. Tesseract OCR engine version 5.5.3 and earlier contains a heap corruption vulnerability due to an unchecked index into the unichars vector. This can cause a crash or potentially controlled corruption. Defenders should assess potential exposure a [truncated]
A crafted TESSDATA_INTTEMP component can cause a heap out-of-bounds write of FontInfo structures in Tesseract OCR engine version 5.5.3 and earlier. This vulnerability can lead to heap corruption, crashes, or potentially controlled corruption of FontInfo structures. The issue arises from the GenericVector::read callback in Tesseract, which fails to properly validate the size_used_ field from a .traineddata [truncated]
PatchSiren debrief for CVE-2026-88050, a MEDIUM-severity vulnerability in Tesseract OCR engine versions 5.5.3 and earlier. The vulnerability occurs in RecodedCharID::DeSerialize, which accepts negative code_ values from a crafted .traineddata recoder component, leading to an out-of-bounds bit write and potential wild-address crashes or allocation failures. Defenders should assess exposure and prioritize p [truncated]
A heap out-of-bounds write vulnerability exists in Tesseract OCR engine versions 5.5.3 and earlier. The issue arises from insufficient bounds checking in the NetworkIO class, specifically in the WriteTimeStepPart and AddTimeStepPart functions. This vulnerability can lead to heap corruption, crashes, or potentially controlled corruption during the first recognition step on the default LSTM engine.
A vulnerability in Tesseract, an open-source OCR engine, allows for a heap out-of-bounds write and read due to improper validation of deserialized layer scalars in the FullyConnected::DeSerialize function. This issue, present in version 5.5.3 and earlier, can lead to heap corruption, crashes, information disclosure, or potentially controlled corruption. The vulnerability is caused by a crafted .traineddat [truncated]
A buffer overflow vulnerability exists in Tesseract's Classify::ReadNormProtos function, which can lead to stack corruption, denial of service, and potentially control-flow hijacking. The vulnerability affects Tesseract version 5.5.3 and earlier. This issue arises from the function's use of std::istream::operator>>(char*) to extract a whitespace-delimited token into a fixed 61-byte stack buffer without se [truncated]
A crafted .traineddata model can cause a heap out-of-bounds read and process crash in Tesseract OCR engine versions before 5.5.3. This issue arises when loading a specially crafted .traineddata model through TessBaseAPI::Init, which can lead to SquishedDawg::read_squished_dawg in src/dict/dawg.cpp accepting an unterminated forward-edge run. As a result, SquishedDawg::Load calls num_forward_edges(0) and la [truncated]
CVE-2026-73066 is a vulnerability in the Tesseract open-source OCR engine, which could lead to a heap out-of-bounds write during OCR recognition. The issue is caused by an unchecked signed integer multiplication in Convolve::DeSerialize in src/lstm/convolve.cpp, which can wrap the convolution output-channel count, undersizing the forward-pass output buffer. The issue is fixed in version 5.5.3. Defenders r [truncated]
