https://cve.radiocsirt.org Contains only the most 10 recent entries. http://www.rssboard.org/rss-specification python-feedgen en Wed, 07 Oct 2026 19:44:29 +0000 https://cve.radiocsirt.org/vuln/certfr-2025-avi-0967 certfr-2025-avi-0967 https://cve.radiocsirt.org/vuln/certfr-2025-avi-0967 https://cve.radiocsirt.org/vuln/euvd-2026-252185 EUVD-2026-252185 https://cve.radiocsirt.org/vuln/euvd-2026-252185 https://cve.radiocsirt.org/vuln/fkie_cve-2025-58446 <p>xgrammar is an open-source library for efficient, flexible, and portable structured generation. A grammar optimizer introduced in 0.1.23 processes large grammars (&gt;100k characters) at very low rates, and can be used for DOS of model providers. This issue is fixed in version 0.1.24.</p> <p>xgrammar is an open-source library for efficient, flexible, and portable structured generation. A grammar optimizer introduced in 0.1.23 processes large grammars (&gt;100k characters) at very low rates, and can be used for DOS of model providers. This issue is fixed in version 0.1.24.</p> https://cve.radiocsirt.org/vuln/fkie_cve-2025-58446 https://cve.radiocsirt.org/vuln/ghsa-9q5r-wfvf-rr7f <p><strong>Affected:</strong> PyPI: xgrammar</p> <p>### Summary Provided grammar, would fit in a context window of most of the models, but takes minutes to process in 0.1.23. In testing with 0.1.16 the parser worked fine so this seems to be a regression caused by Earley parser.</p> <p>### Details</p> <p>Full reproducer provider in the POC section. The resulting grammar is around 70k tokens, and the grammar parsing itself (with the models I checked) was significantly longer than LLM processing itself, meaning this can be used to DOS model providers.</p> <p>### Patch</p> <p>This problem is caused by the grammar optimizer introduced in v0.1.23 being too slow. It only happens for very large grammars (&gt;100k characters), like the below one. v0.1.24 solved this problem by optimizing the speed of the grammar optimizer and disable some slow optimization for large grammars.</p> <p>Thanks to @Seven-Streams</p> <p>### PoC ``` import string import random</p> <p>def enum_schema(size=10000,str_len=10): enum = {&#34;enum&#34;: [&#34;&#34;.join(random.choices(string.ascii_uppercase, k=str_len)) for _ in range(size)]} schema = { &#34;definitions&#34;: { &#34;colorEnum&#34;: enum }, &#34;type&#34;: &#34;object&#34;, &#34;properties&#34;: { &#34;color1&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color2&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color3&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color4&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34;…</p> <p><strong>Affected:</strong> PyPI: xgrammar</p> <p>### Summary Provided grammar, would fit in a context window of most of the models, but takes minutes to process in 0.1.23. In testing with 0.1.16 the parser worked fine so this seems to be a regression caused by Earley parser.</p> <p>### Details</p> <p>Full reproducer provider in the POC section. The resulting grammar is around 70k tokens, and the grammar parsing itself (with the models I checked) was significantly longer than LLM processing itself, meaning this can be used to DOS model providers.</p> <p>### Patch</p> <p>This problem is caused by the grammar optimizer introduced in v0.1.23 being too slow. It only happens for very large grammars (&gt;100k characters), like the below one. v0.1.24 solved this problem by optimizing the speed of the grammar optimizer and disable some slow optimization for large grammars.</p> <p>Thanks to @Seven-Streams</p> <p>### PoC ``` import string import random</p> <p>def enum_schema(size=10000,str_len=10): enum = {&#34;enum&#34;: [&#34;&#34;.join(random.choices(string.ascii_uppercase, k=str_len)) for _ in range(size)]} schema = { &#34;definitions&#34;: { &#34;colorEnum&#34;: enum }, &#34;type&#34;: &#34;object&#34;, &#34;properties&#34;: { &#34;color1&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color2&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color3&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color4&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34;…</p> https://cve.radiocsirt.org/vuln/ghsa-9q5r-wfvf-rr7f https://cve.radiocsirt.org/vuln/pysec-2026-2055 <p><strong>Affected:</strong> PyPI: xgrammar</p> <p>### Summary Provided grammar, would fit in a context window of most of the models, but takes minutes to process in 0.1.23. In testing with 0.1.16 the parser worked fine so this seems to be a regression caused by Earley parser.</p> <p>### Details</p> <p>Full reproducer provider in the POC section. The resulting grammar is around 70k tokens, and the grammar parsing itself (with the models I checked) was significantly longer than LLM processing itself, meaning this can be used to DOS model providers.</p> <p>### Patch</p> <p>This problem is caused by the grammar optimizer introduced in v0.1.23 being too slow. It only happens for very large grammars (&gt;100k characters), like the below one. v0.1.24 solved this problem by optimizing the speed of the grammar optimizer and disable some slow optimization for large grammars.</p> <p>Thanks to @Seven-Streams</p> <p>### PoC ``` import string import random</p> <p>def enum_schema(size=10000,str_len=10): enum = {&#34;enum&#34;: [&#34;&#34;.join(random.choices(string.ascii_uppercase, k=str_len)) for _ in range(size)]} schema = { &#34;definitions&#34;: { &#34;colorEnum&#34;: enum }, &#34;type&#34;: &#34;object&#34;, &#34;properties&#34;: { &#34;color1&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color2&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color3&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color4&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34;…</p> <p><strong>Affected:</strong> PyPI: xgrammar</p> <p>### Summary Provided grammar, would fit in a context window of most of the models, but takes minutes to process in 0.1.23. In testing with 0.1.16 the parser worked fine so this seems to be a regression caused by Earley parser.</p> <p>### Details</p> <p>Full reproducer provider in the POC section. The resulting grammar is around 70k tokens, and the grammar parsing itself (with the models I checked) was significantly longer than LLM processing itself, meaning this can be used to DOS model providers.</p> <p>### Patch</p> <p>This problem is caused by the grammar optimizer introduced in v0.1.23 being too slow. It only happens for very large grammars (&gt;100k characters), like the below one. v0.1.24 solved this problem by optimizing the speed of the grammar optimizer and disable some slow optimization for large grammars.</p> <p>Thanks to @Seven-Streams</p> <p>### PoC ``` import string import random</p> <p>def enum_schema(size=10000,str_len=10): enum = {&#34;enum&#34;: [&#34;&#34;.join(random.choices(string.ascii_uppercase, k=str_len)) for _ in range(size)]} schema = { &#34;definitions&#34;: { &#34;colorEnum&#34;: enum }, &#34;type&#34;: &#34;object&#34;, &#34;properties&#34;: { &#34;color1&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color2&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color3&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34; }, &#34;color4&#34;: { &#34;$ref&#34;: &#34;#/definitions/colorEnum&#34;…</p> https://cve.radiocsirt.org/vuln/pysec-2026-2055