Unlearning Isn'T Deletion: Investigating Reversibility of Machine Unlearning in LLMs

Unlearning in large language models LLMs is intended to remove the influence of specific data, yet current evaluations rely heavily on token-level metrics such as accuracy and perplexity. We show that these metrics can be misleading: models often appear to forget, but their original behavior can ...

MTSA: Multi-Turn Safety Alignment for LLMs through Multi-Round Red-Teaming

Whitepaper called MTSA: Multi-Turn Safety Alignment For LLMs Through Multi-Round Red-Teaming...

PT-2025-22536 · Unknown · Nexus Series +2

Name of the Vulnerable Software and Affected Versions: ASPECT-Enterprise versions through 3. NEXUS Series versions through 3. MATRIX Series versions through 3. Description: The issue is related to large content vulnerabilities in ASPECT, which can expose a device to disk overutilization on a syst...

PT-2025-22546 · Unknown · Nexus Series +2

Name of the Vulnerable Software and Affected Versions: ASPECT-Enterprise versions through 3. NEXUS Series versions through 3. MATRIX Series versions through 3. Description: The issue is related to large content vulnerabilities in ASPECT, which can expose a device to disk overutilization on a syst...

DuFFin: a Dual-Level Fingerprinting Framework for LLMs IP Protection

Whitepaper called DuFFin: A Dual-Level Fingerprinting Framework For LLMs IP Protection...

CVE-2002-2075

ICQ 2001a and 2002b allows remote attackers to cause a denial of service memory consumption and hang via a contact message with a large contacts number...

CVE-2008-5421

The SSL web administration service in NetWin SmsGate 1.1n and earlier allows remote attackers to cause a denial of service hang via 1 a large integer in the Content-Length HTTP header; 2 an invalid value in the Content-Length HTTP header, as demonstrated by a negative integer; or 3 a missing...

CVE-2003-0366

lyskom-server 2.0.7 and earlier allows unauthenticated users to cause a denial of service CPU consumption via a large query...

CVE-1999-0465

Remote attackers can crash Lynx and Internet Explorer using an IMG tag with a large width parameter...

DRUPAL-CONTRIB-2025-068

The Admin Audit Trail module tracks logs of specific events that you'd like to review. When the submodule Admin Audit Trail: User Authentication is enabled, it logs user authentication events login, logout, and password reset requests. The module does not sufficiently limit some large values befo...

AZL-61977 CVE-2024-23337 affecting package jq for versions less than 1.7.1-3

jq is a command-line JSON processor. In versions up to and including 1.7.1, an integer overflow arises when assigning value using an index of 2147483647, the signed integer limit. This causes a denial of service. Commit de21386681c0df0104a99d9d09db23a9b2a78b1e contains a patch for the issue...

SafeKey: Amplifying Aha-Moment Insights for Safety Reasoning

Large Reasoning Models LRMs introduce a new generation paradigm of explicitly reasoning before answering, leading to remarkable improvements in complex tasks. However, they pose great safety risks against harmful queries and adversarial attacks. While recent mainstream safety efforts on LRMs,...

Leveraging Large Language Models for Command Injection Vulnerability Analysis in Python: an Empirical Study on Popular Open-Source Projects

Command injection vulnerabilities are a significant security threat in dynamic languages like Python, particularly in widely used open-source projects where security issues can have extensive impact. With the proven effectiveness of Large Language ModelsLLMs in code-related tasks, such as testing...

unbound: Unbounded name compression could lead to Denial of Service

A flaw was found in Unbound which can lead to degraded performance and an eventual denial of service when handling replies with very large RRsets that require name compression to be applied. Versions prior to 1.21.1 do not have a hard limit on the number of name compression calculations that...

DEBIAN-CVE-2025-37931

In the Linux kernel, the following vulnerability has been resolved: btrfs: adjust subpage bit start based on sectorsize When running machines with 64k page size and a 16k nodesize we started seeing tree log corruption in production. This turned out to be because we were not writing out dirty bloc...

SUSE-SU-2025:20330-1 Security update for python-h11, python-httpcore

This update for python-h11, python-httpcore fixes the following issues: python-h11: - Update 0.16.0: CVE-2025-43859: Fixed accepting of malformed Chunked-Encoding bodies bsc1241872 - 0.15.0: Reject Content-Lengths = 1 zettabyte 1 billion terabytes early, without attempting to parse the integer 18...

Security update for python-h11, python-httpcore

This update for python-h11, python-httpcore fixes the following issues: python-h11: - Update 0.16.0: CVE-2025-43859: Fixed accepting of malformed Chunked-Encoding bodies bsc1241872 - 0.15.0: Reject Content-Lengths = 1 zettabyte 1 billion terabytes early, without attempting to parse the integer...

libtasn1: Inefficient DER Decoding in libtasn1 Leading to Potential Remote DoS

A flaw in libtasn1 causes inefficient handling of specific certificate data. When processing a large number of elements in a certificate, libtasn1 takes much longer than expected, which can slow down or even crash the system. This flaw allows an attacker to send a specially crafted certificate,...

Integer Overflow or Wraparound

Overview Affected versions of this package are vulnerable to Integer Overflow or Wraparound via the archivereadsupportformatwarc.c process. An attacker can cause memory corruption or unpredictable program behavior by supplying a specially crafted WARC file with an excessively large content length...

From Nuclear Safety to LLM Security: Applying Non-Probabilistic Risk Management Strategies to Build Safe and Secure LLM-Powered Systems

Large language models LLMs offer unprecedented and growing capabilities, but also introduce complex safety and security challenges that resist conventional risk management. While conventional probabilistic risk analysis PRA requires exhaustive risk enumeration and quantification, the novelty and...