CVE-2026-35502

Deserialization of untrusted data for some IntelR Extension for PyTorch before version 2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with an unauthenticated user combined with a low complexity attack may enable escalation of privilege...

CVE-2026-35502

CVE-2026-35502 describes a deserialization vulnerability in Intel(R) Extension for PyTorch prior to 2.8.0, affecting Ring 3: User Applications. An unprivileged, unauthenticated user could achieve local privilege escalation with low impact to confidentiality, integrity, and availability. The issue...

EUVD-2026-56519

Protection mechanism failure for some IntelR oneCCL Bindings for PyTorch before version v2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. Th...

CVE-2026-24693

Protection mechanism failure for some IntelR oneCCL Bindings for PyTorch before version v2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. Th...

CVE-2026-24693

Protection mechanism failure for some IntelR oneCCL Bindings for PyTorch before version v2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. Th...

CVE-2026-24693

Intel oneAPI/oneCCL Bindings for PyTorch (before v2.8.0) is reported to have a Protection mechanism failure that could allow a local escalation of privilege when a low-complexity, attacker-controlled scenario involves a user with elevated privileges and passive user interaction. Affected componen...

CVE-2026-21387

Protection mechanism failure for some IntelR LLM Library for PyTorch within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. This result may potentially...

CVE-2026-21387

Protection mechanism failure for some IntelR LLM Library for PyTorch within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. This result may potentially...

CVE-2026-21387

CVE-2026-21387 affects Intel(R) LLM Library for PyTorch reported to have a protection mechanism failure in Ring 3. The vulnerability may allow an escalation of privilege via a local attacker with a privileged user and low attack complexity, with passive user interaction. Impact is described as hi...

PT-2026-70328

Protection mechanism failure for some IntelR oneCCL Bindings for PyTorch before version v2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. Th...

PT-2026-70342

Deserialization of untrusted data for some IntelR Extension for PyTorch before version 2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with an unauthenticated user combined with a low complexity attack may enable escalation of privilege...

CVE-2026-24693: Protection Mechanism Failure

Protection mechanism failure for some IntelR oneCCL Bindings for PyTorch before version v2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. Th...

CVE-2026-35502: Deserialization of Untrusted Data

Deserialization of untrusted data for some IntelR Extension for PyTorch before version 2.8.0 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with an unauthenticated user combined with a low complexity attack may enable escalation of privilege...

Intel® Extension for PyTorch Software Advisory

Summary: A potential security vulnerability for some Intel® Extension for PyTorch software may allow escalation of privilege. Intel is not releasing updates to mitigate this potential vulnerability and has issued a Product Discontinuation Notice for the Intel® Extension for PyTorch software...

Intel® LLM Library for PyTorch Software Advisory

Summary: A potential security vulnerability for some Intel® LLM Library for PyTorch software may allow escalation of privilege. Intel is not releasing updates to mitigate this potential vulnerability and has issued a Product Discontinuation Notice for the Intel® LLM Library for PyTorch software...

CVE-2026-21387: Protection Mechanism Failure

Protection mechanism failure for some IntelR LLM Library for PyTorch within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. This result may potentially...

PYSEC-2026-3634 Keras: TorchModuleWrapper can deserialize unsafe PyTorch pickle data

A vulnerability in keras-team/keras version 3.15.0 allows unsafe deserialization of attacker-controlled PyTorch pickle data through the public keras.layers.TorchModuleWrapper.fromconfig method. This method invokes torch.load..., weightsonly=False without requiring an explicit unsafe opt-in, such ...

PT-2026-76467

A vulnerability in keras-team/keras version 3.15.0 allows unsafe deserialization of attacker-controlled PyTorch pickle data through the public keras.layers.TorchModuleWrapper.from config method. This method invokes torch.load..., weights only=False without requiring an explicit unsafe opt-in, suc...

CVE-2026-31221

A flaw was found in PyTorch-Lightning. This vulnerability, categorized as insecure deserialization CWE-502, exists in the checkpoint loading mechanism. A remote attacker can exploit this by providing a maliciously crafted checkpoint file, which, when loaded by the LightningModule.loadfromcheckpoi...

Dual-Node NVIDIA DGX Spark over Tailscale: A Remote-Access Testbed for Distributed LLM Training and Cyber-Threat-Intelligence Fine-Tuning

Compact AI systems make local language-model experimentation increasingly accessible, yet practical evidence for multi-node training on desktop-class accelerators remains limited. This report presents a proof-of-concept deployment of distributed NanoChat pretraining across two NVIDIA DGX Spark...