CVE-2026-88054

A commercial quality OCR engine originally developed at HP between 1985 and 1995. In 1995, this engine was among the top 3 evaluated by UNLV. It was open-sourced by HP and UNLV in 2005.

Security Fix(es):

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, Classify::ReadNormProtos in src/classify/normmatch.cpp parses the NORMPROTO component of a .traineddata file and uses std::istream::operator>>(char*) to extract a whitespace-delimited token into a fixed 61-byte stack buffer without setting a stream width. The 100-byte line buffer can carry a token of up to 99 characters, so a token longer than 60 characters writes up to 39 attacker-controlled bytes past the buffer during TessBaseAPI::Init of the legacy engine, causing stack corruption, denial of service, and potentially control-flow hijacking on affected standard-library implementations. Builds using Apple's libc++ C++20 bounded array overload are incidentally protected, while typical libstdc++ builds remain affected. No fixed release is available as of this review.(CVE-2026-88047)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, FullyConnected::DeSerialize in src/lstm/fullyconnected.cpp does not validate the deserialized layer scalars ni_ and no_ against the weight-matrix dimensions. During FullyConnected::Forward, MatrixDotVector in src/lstm/weightmatrix.cpp writes w.dim1() results into temp_line, which is sized from no_, and reads w.dim2() minus one inputs from curr_input, which is sized from ni_. A crafted .traineddata NT_SOFTMAX layer can therefore use inconsistent dimensions to cause a heap out-of-bounds write and read on the default LSTM engine, resulting in heap corruption, a crash, information disclosure, or potentially controlled corruption. No fixed release is available as of this review.(CVE-2026-88048)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, prior .traineddata hardening added bounds checks to NetworkIO::CopyTimeStepGeneral and NetworkIO::Randomize in src/lstm/networkio.cpp but left NetworkIO::WriteTimeStepPart and NetworkIO::AddTimeStepPart unchecked. In LSTM::Forward in src/lstm/lstm.cpp, source_ is sized from the independently deserialized na_ field while the WriteTimeStepPart count is ns_, which comes from the CI gate WeightMatrix dim1() value. A crafted NT_LSTM layer can make ns_ much larger than na_, causing a heap out-of-bounds write during the first recognition step on the default LSTM engine and resulting in heap corruption, a crash, or potentially controlled corruption. No fixed release is available as of this review.(CVE-2026-88049)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, RecodedCharID::DeSerialize in src/ccutil/unicharcompress.h validates length_ but accepts negative code_ values from a crafted .traineddata recoder component. UnicharCompress::ComputeCodeRange in src/ccutil/unicharcompress.cpp can consequently produce code_range_ equal to zero, after which SetupDecoder indexes is_valid_start_ with the negative code on a size-zero vector. The resulting out-of-bounds bit write uses a large wrapped index and reliably causes a wild-address crash or allocation failure on the default LSTM engine. No fixed release is available as of this review.(CVE-2026-88050)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, the callback form of GenericVector::read in src/ccutil/genericvector.h reads the independent int32 fields reserved and size_used_ from a .traineddata model without a cap or an invariant check. reserve(reserved) allocates the backing array, but the callback loop writes size_used_ elements. A crafted TESSDATA_INTTEMP component with version_id 4 or later can therefore set reserved to a small value and size_used_ to a large value when fontinfo_table_.read(fp, read_info) is called from src/classify/intproto.cpp, causing a heap out-of-bounds write of FontInfo structures, heap corruption, a crash, or potentially controlled…

A commercial quality OCR engine originally developed at HP between 1985 and 1995. In 1995, this engine was among the top 3 evaluated by UNLV. It was open-sourced by HP and UNLV in 2005.

Security Fix(es):

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, Classify::ReadNormProtos in src/classify/normmatch.cpp parses the NORMPROTO component of a .traineddata file and uses std::istream::operator>>(char*) to extract a whitespace-delimited token into a fixed 61-byte stack buffer without setting a stream width. The 100-byte line buffer can carry a token of up to 99 characters, so a token longer than 60 characters writes up to 39 attacker-controlled bytes past the buffer during TessBaseAPI::Init of the legacy engine, causing stack corruption, denial of service, and potentially control-flow hijacking on affected standard-library implementations. Builds using Apple's libc++ C++20 bounded array overload are incidentally protected, while typical libstdc++ builds remain affected. No fixed release is available as of this review.(CVE-2026-88047)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, FullyConnected::DeSerialize in src/lstm/fullyconnected.cpp does not validate the deserialized layer scalars ni_ and no_ against the weight-matrix dimensions. During FullyConnected::Forward, MatrixDotVector in src/lstm/weightmatrix.cpp writes w.dim1() results into temp_line, which is sized from no_, and reads w.dim2() minus one inputs from curr_input, which is sized from ni_. A crafted .traineddata NT_SOFTMAX layer can therefore use inconsistent dimensions to cause a heap out-of-bounds write and read on the default LSTM engine, resulting in heap corruption, a crash, information disclosure, or potentially controlled corruption. No fixed release is available as of this review.(CVE-2026-88048)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, prior .traineddata hardening added bounds checks to NetworkIO::CopyTimeStepGeneral and NetworkIO::Randomize in src/lstm/networkio.cpp but left NetworkIO::WriteTimeStepPart and NetworkIO::AddTimeStepPart unchecked. In LSTM::Forward in src/lstm/lstm.cpp, source_ is sized from the independently deserialized na_ field while the WriteTimeStepPart count is ns_, which comes from the CI gate WeightMatrix dim1() value. A crafted NT_LSTM layer can make ns_ much larger than na_, causing a heap out-of-bounds write during the first recognition step on the default LSTM engine and resulting in heap corruption, a crash, or potentially controlled corruption. No fixed release is available as of this review.(CVE-2026-88049)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, RecodedCharID::DeSerialize in src/ccutil/unicharcompress.h validates length_ but accepts negative code_ values from a crafted .traineddata recoder component. UnicharCompress::ComputeCodeRange in src/ccutil/unicharcompress.cpp can consequently produce code_range_ equal to zero, after which SetupDecoder indexes is_valid_start_ with the negative code on a size-zero vector. The resulting out-of-bounds bit write uses a large wrapped index and reliably causes a wild-address crash or allocation failure on the default LSTM engine. No fixed release is available as of this review.(CVE-2026-88050)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, the callback form of GenericVector::read in src/ccutil/genericvector.h reads the independent int32 fields reserved and size_used_ from a .traineddata model without a cap or an invariant check. reserve(reserved) allocates the backing array, but the callback loop writes size_used_ elements. A crafted TESSDATA_INTTEMP component with version_id 4 or later can therefore set reserved to a small value and size_used_ to a large value when fontinfo_table_.read(fp, read_info) is called from src/classify/intproto.cpp, causing a heap out-of-bounds write of FontInfo structures, heap corruption, a crash, or potentially controlled…

A commercial quality OCR engine originally developed at HP between 1985 and 1995. In 1995, this engine was among the top 3 evaluated by UNLV. It was open-sourced by HP and UNLV in 2005.

Security Fix(es):

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, Classify::ReadNormProtos in src/classify/normmatch.cpp parses the NORMPROTO component of a .traineddata file and uses std::istream::operator>>(char*) to extract a whitespace-delimited token into a fixed 61-byte stack buffer without setting a stream width. The 100-byte line buffer can carry a token of up to 99 characters, so a token longer than 60 characters writes up to 39 attacker-controlled bytes past the buffer during TessBaseAPI::Init of the legacy engine, causing stack corruption, denial of service, and potentially control-flow hijacking on affected standard-library implementations. Builds using Apple's libc++ C++20 bounded array overload are incidentally protected, while typical libstdc++ builds remain affected. No fixed release is available as of this review.(CVE-2026-88047)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, FullyConnected::DeSerialize in src/lstm/fullyconnected.cpp does not validate the deserialized layer scalars ni_ and no_ against the weight-matrix dimensions. During FullyConnected::Forward, MatrixDotVector in src/lstm/weightmatrix.cpp writes w.dim1() results into temp_line, which is sized from no_, and reads w.dim2() minus one inputs from curr_input, which is sized from ni_. A crafted .traineddata NT_SOFTMAX layer can therefore use inconsistent dimensions to cause a heap out-of-bounds write and read on the default LSTM engine, resulting in heap corruption, a crash, information disclosure, or potentially controlled corruption. No fixed release is available as of this review.(CVE-2026-88048)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, prior .traineddata hardening added bounds checks to NetworkIO::CopyTimeStepGeneral and NetworkIO::Randomize in src/lstm/networkio.cpp but left NetworkIO::WriteTimeStepPart and NetworkIO::AddTimeStepPart unchecked. In LSTM::Forward in src/lstm/lstm.cpp, source_ is sized from the independently deserialized na_ field while the WriteTimeStepPart count is ns_, which comes from the CI gate WeightMatrix dim1() value. A crafted NT_LSTM layer can make ns_ much larger than na_, causing a heap out-of-bounds write during the first recognition step on the default LSTM engine and resulting in heap corruption, a crash, or potentially controlled corruption. No fixed release is available as of this review.(CVE-2026-88049)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, RecodedCharID::DeSerialize in src/ccutil/unicharcompress.h validates length_ but accepts negative code_ values from a crafted .traineddata recoder component. UnicharCompress::ComputeCodeRange in src/ccutil/unicharcompress.cpp can consequently produce code_range_ equal to zero, after which SetupDecoder indexes is_valid_start_ with the negative code on a size-zero vector. The resulting out-of-bounds bit write uses a large wrapped index and reliably causes a wild-address crash or allocation failure on the default LSTM engine. No fixed release is available as of this review.(CVE-2026-88050)

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, the callback form of GenericVector::read in src/ccutil/genericvector.h reads the independent int32 fields reserved and size_used_ from a .traineddata model without a cap or an invariant check. reserve(reserved) allocates the backing array, but the callback loop writes size_used_ elements. A crafted TESSDATA_INTTEMP component with version_id 4 or later can therefore set reserved to a small value and size_used_ to a large value when fontinfo_table_.read(fp, read_info) is called from src/classify/intproto.cpp, causing a heap out-of-bounds write of FontInfo structures, heap corruption, a crash, or potentially controlled…

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, Plumbing::DeSerialize in src/lstm/plumbing.cpp rejects excessively large network stacks but accepts a zero-length stack for NT_SERIES, NT_PARALLEL, or NT_REVERSED layers in a crafted .traineddata model. During LSTMRecognizer initialization in src/lstm/lstmrecognizer.cpp, CacheXScaleFactor(XScaleFactor()) reaches Series::CacheXScaleFactor in src/lstm/series.cpp, which dereferences stack_[0] on the empty vector and invokes a virtual method through an invalid Network pointer. This causes a deterministic crash and denial of service at model load. No fixed release is available as of this review.

This is a security update. It fixes nine vulnerabilities in tesseract, the OCR (text recognition) engine. All nine can be triggered only by feeding tesseract a maliciously crafted input file: either a specially crafted recognition/language data file (.traineddata) or a crafted word-list file. Depending on the specific flaw, this can cause a crash (denial of service) or memory corruption. CVE-2026-73067: out-of-bounds read when loading a crafted word-list file. CVE-2026-73066: out-of-bounds write when loading a crafted recognition model file. CVE-2026-88047: stack buffer overflow when loading a crafted language-normalisation file. CVE-2026-88048: out-of-bounds read/write from a crafted neural-network layer in a recognition model file. CVE-2026-88049: out-of-bounds write from a crafted neural-network layer in a recognition model file. CVE-2026-88050: out-of-bounds write from invalid character-encoding values in a crafted recognition model file. CVE-2026-88051: out-of-bounds write when loading a malformed internal data structure from a crafted recognition model file. CVE-2026-88052: out-of-bounds write when loading a crafted character-set file. CVE-2026-88053: out-of-bounds write when loading a crafted legacy recognition template file. CVE-2026-88054: crash (denial of service) when loading a crafted recognition model with an empty internal network. The tesseract package is updated to version 5.5.3, which on its own fixes CVE-2026-73067 and CVE-2026-73066. The remaining seven issues (CVE-2026-88047 to CVE-2026-88054) are not yet fixed in any upstream release, so nine patches taken from upstream's main development branch are also applied. Reported by Tristan Madani (CVE-2026-88047) and Zhixi "Jace" Sun (CVE-2026-88048 to CVE-2026-88054); see the individual advisories below for full credits, including CVE-2026-73067 and CVE-2026-73066.

These are all security issues fixed in the libtesseract5-5.5.3-2.1 package on the GA media of openSUSE Tumbleweed.

This update for tesseract-ocr fixes the following issues:

Changes in tesseract-ocr:

• Update to version 5.5.3 (sync with Factory):
  • CVE-2026-73067: heap out-of-bounds read in SquishedDawg on crafted model (boo#1275623)
  • CVE-2026-88047: stack buffer overflow in Classify::ReadNormProtos on crafted traineddata (boo#1280925)
  • CVE-2026-88048: heap out-of-bounds write/read in FullyConnected::Forward via dimension mismatch (boo#1280929)
  • CVE-2026-88049: heap out-of-bounds write in LSTM::Forward via na_/gate-matrix dimension mismatch (boo#1280930)
  • CVE-2026-88050: out-of-bounds write in UnicharCompress via unvalidated recoder code values (boo#1280931)
  • CVE-2026-88051: heap out-of-bounds write in GenericVector<T>::read via reserved/size_used_ mismatch (boo#1280932)
  • CVE-2026-88052: heap out-of-bounds write in UNICHARSET::load_via_fgets via count/insert desynchronization (boo#1280933)
  • CVE-2026-88053: heap out-of-bounds write in Classify::ReadIntTemplates via unvalidated counts in crafted traineddata (boo#1280934)
  • CVE-2026-88054: denial of service via empty-stack dereference at model load (boo#1280935)

It was discovered that Tesseract incorrectly handled crafted .traineddata models. An attacker could possibly use this issue to cause Tesseract to crash, resulting in a denial of service. (CVE-2026-73067)

It was discovered that Tesseract did not properly validate dimensions in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-73066)

It was discovered that Tesseract did not properly limit token lengths in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-88047)

It was discovered that Tesseract did not properly validate layer dimensions in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-88048)

It was discovered that Tesseract did not properly validate layer sizes in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-88049)

It was discovered that Tesseract incorrectly handled negative character codes in crafted .traineddata models. An attacker could possibly use this issue to cause Tesseract to crash, resulting in a denial of service. (CVE-2026-88050)

It was discovered that Tesseract did not properly validate vector sizes in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-88051)

It was discovered that Tesseract did not properly validate character IDs in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-88052)

It was discovered that Tesseract did not properly validate classifier counts in crafted .traineddata models. An attacker could possibly use this issue to execute arbitrary code. (CVE-2026-88053)

It was discovered that Tesseract incorrectly handled empty network layers in crafted .traineddata models. An attacker could possibly use this issue to cause Tesseract to crash, resulting in a denial of service. (CVE-2026-88054)