Breaking the Code: Security Assessment of AI Code Agents through Systematic Jailbreaking Attacks

Code-capable large language model LLM agents are increasingly embedded into software engineering workflows where they can read, write, and execute code, raising the stakes of safety-bypass "jailbreak" attacks beyond text-only settings. Prior evaluations emphasize refusal or harmful-text detection...

ODSafeManager#allowSAFE() cannot be executed either by the proxy contract or any other address.

Lines of code Vulnerability details Impact "According to the GEB framework, the proxy contracts ODProxy are designed to interact with the Safe Manager ODSafeManager through the Proxy Action contract BasicActions. The pivotal function, allowSAFE, is responsible for granting an address the capabili...

Code injection

An incorrect setting of UXN bits within mmuflagstos1pteattr lead to privileged executable pages being mapped as executable from an unprivileged context. This can be leveraged by an attacker to bypass executability restrictions of kernel-mode pages from user-mode. An incorrect setting of PXN bits...

CVE-2021-22566

The CVE concerns the Fuchsia kernel. An incorrect setting of UXN and PXN bits in mmu_flags_to_s1_pte_attr can cause privileged executable kernel-mode pages to be mapped as executable from user mode, and unprivileged pages to be treated as executable from kernel mode. This misuse can bypass execut...