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.

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 corruption. No fixed release is available as of this review.

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.

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.

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.

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.

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, UNICHARSET::load_via_fgets in src/ccutil/unicharset.cpp trusts the declared unichar count as a loop bound and uses id as an unchecked index into the unichars vector. unichar_insert_backwards_compatible can leave the vector unchanged for an empty, duplicate, or already-encodable representation, causing id to become larger than unichars.size(). Subsequent set_* calls and the write to unichars[id].properties.enabled then write UNICHAR_PROPERTIES beyond the vector during initialization in both the default LSTM and legacy engines, causing heap corruption, a crash, or potentially controlled corruption. No fixed release is available as of this review.

Tesseract is an open source OCR engine. In version 5.5.3 and earlier, Classify::ReadIntTemplates in src/classify/intproto.cpp reads NumClassPruners, NumClasses, and NumProtoSets from the TESSDATA_INTTEMP component of a crafted .traineddata file and uses those values as loop bounds without validating them against MAX_NUM_CLASS_PRUNERS, MAX_NUM_CLASSES, and MAX_NUM_PROTO_SETS. The loops store heap pointers into fixed-capacity ClassPruners and ProtoSets arrays in INT_TEMPLATES_STRUCT and INT_CLASS_STRUCT, so an oversized count causes heap out-of-bounds pointer writes during legacy-classifier initialization before OCR begins, resulting in heap corruption, a crash, or potentially controlled corruption. No fixed release is available as of this review.