17 matches found
Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection
Prompt injection against LLM agents becomes much stronger when the injected instruction is wrapped in the model's own chat template. A forged template marker such as can reach the model either as a single reserved control token or as a sequence of ordinary subword tokens. The two decode to exactl...
Transformers Regular Expression Denial of Service (ReDoS) vulnerability
A Regular Expression Denial of Service ReDoS vulnerability was identified in the huggingface/transformers library, specifically in the file tokenizationgptneoxjapanese.py of the GPT-NeoX-Japanese model. The vulnerability occurs in the SubWordJapaneseTokenizer class, where regular expressions...
HSEC-2025-0001 Subword division operations may produce incorrect results
Subword division operations may produce incorrect results Arithmetic operations may produce incorrect results when compiled with optimizations. For the most part, this bug only causes availability and data integrity issues. However, in some circumstances, it may result in other, more complicated...
URL2Graph++: Unified Semantic-Structural-Character Learning for Malicious URL Detection
Malicious URL detection remains a major challenge in cybersecurity, primarily due to two factors: 1 the exponential growth of the Internet has led to an immense diversity of URLs, making generalized detection increasingly difficult; and 2 attackers are increasingly employing sophisticated...
Regular Expression Denial of Service (ReDoS)
Overview transformers is a State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow Affected versions of this package are vulnerable to Regular Expression Denial of Service ReDoS in the tokenizationgptneoxjapanese.py file, within the SubWordJapaneseTokenizer class. The regex is designed ...
PT-2025-11651
Subword division operations may produce incorrect results Arithmetic operations may produce incorrect results when compiled with optimizations. For the most part, this bug only causes availability and data integrity issues. However, in some circumstances, it may result in other, more complicated...
SUSE CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
DEBIAN-CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191: Observable Timing Discrepancy
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
CVE-2024-45191
The Matrix libolm library (through version 3.2.16 ) contains an AES implementation vulnerable to cache-timing attacks . This is caused by the use of S-boxes in software utilizing a lookup table for the SubWord step. This vulnerability specifically affects products no longer supported by the maint...
CVE-2024-45191
An issue was discovered in Matrix libolm through 3.2.16. The AES implementation is vulnerable to cache-timing attacks due to use of S-boxes. This is related to software that uses a lookup table for the SubWord step. This refers to the libolm implementation of Olm. NOTE: This vulnerability only...
poppler/pdf_fuzzer: Use-of-uninitialized-value in subWord
Project: https://anongit.freedesktop.org/git/poppler/poppler.git https://gitlab.freedesktop.org/poppler/poppler.git https://gitlab.freedesktop.org/ceyhunalp/poppler.git Detailed report: https://oss-fuzz.com/testcase?key=6332584257650688 Project: poppler Fuzzer: libFuzzerpopplerpdffuzzer Fuzz targ...