Aggregator

A vulnerability classified as problematic has been found in Wireshark up to 4.4.13/4.6.3. This affects an unknown function of the component USB HID Protocol Dissector. This manipulation causes improperly controlled sequential memory allocation. This vulnerability appears as CVE-2026-3201. The attack may be initiated remotely. There is no available exploit. It is recommended to upgrade the affected component.

A vulnerability described as problematic has been identified in InternationalColorConsortium iccDEV up to 2.3.1.4. The impacted element is an unknown function of the file iccFromCube.cpp. The manipulation results in integer overflow. This vulnerability is reported as CVE-2026-27691. The attack can be launched remotely. No exploit exists. It is advisable to implement a patch to correct this issue.

SentinelOne has unveiled its Singularity Identity portfolio designed to secure the growing population of non-human identities, including AI agents, service accounts, APIs, and workloads. Identity attacks have long been a go-to tactic for nation-state actors and cybercriminals. Most defenses focus on stopping them at authentication and permissions. Attackers continue adapting their tactics to bypass those controls. When a threat actor logs in as a legitimate user and uses approved tools to move laterally or exfiltrate … More →

The post SentinelOne addresses identity risk across endpoints, browsers, and AI workflows appeared first on Help Net Security.

Разработчики Firefox 148 закрыли опасные уязвимости в системе защиты памяти.

根据发表在预印本平台 arXiv 上的一篇论文，AI 总是在战争模拟游戏中推荐核打击，而人类在使用核武器上则有更多顾虑。伦敦国王学院的 Kenneth Payne 让三个主流模型 GPT-5.2、Claude Sonnet 4 和 Gemini 3 Flash 在模拟战争游戏中互相对抗，游戏场景包括激烈的国际对峙，涉及边界争端、稀缺资源争夺以及政权的生存威胁。AI 允许采取从外交抗议、彻底投降到全面核战争等一系列行动。AI 进行了 21 场游戏，329 个回合，生成了 78 万字去描述其决策背后的逻辑。在 95% 的模拟游戏中，AI 模型至少部署了一枚战术核武器。普林斯顿的 Tong Zhao 称，主要大国已在战争模拟中纳入 AI，但目前并不确定 AI 的决策支持在多大程度上纳入实际的军事决策。Payne 认为没人会把核导弹发射井的控制权交给 AI，任由它们做出决定。三个模型的开发商 OpenAI、Anthropic 和 Google 未对该研究置评。

IBM's 2026 X-Force report reveals 44% rise in cyber-attacks on public apps, driven by AI and flaws

Triage is supposed to make things simpler. In a lot of teams, it does the opposite. When you can’t reach a confident verdict early, alerts turn into repeat checks, back-and-forth, and “just escalate it” calls. That cost doesn’t stay inside the SOC; it shows up as missed SLAs, higher cost per case, and more room for real threats to slip through. So where does triage go wrong? Here are five triage

How much time do you spend reviewing alerts that turn out to be harmless?  In many teams, a single alert takes around 30 minutes to investigate. Not because it’s complex, but because you have to pull context from multiple tools before reaching a confident verdict: reputation checks, enrichment, detonation requests, log pivots.  At scale, that […]

The post How SOC Analysts Can Save 28 Minutes Per Alert Review  appeared first on Cyber Security News.

Veza expanded its platform with Veza Access Agents, a set of purpose-built AI agents designed to automate complex identity and access governance tasks for enterprises. Veza also announced advancements to its AI Agent Security product, providing organizations with deeper visibility into agent risks and greater control over third-party AI agents, large language models (LLMs), AI apps, and AI infrastructure (MCP). As enterprises adopt agentic AI to drive business efficiency, the volume and complexity of identity-based … More →

The post Veza expands platform with AI Access Agents for enterprise identity governance appeared first on Help Net Security.

Alisa Viejo, CA, United States, 25th February 2026, CyberNewswire

Почему бесплатные данные для обучения ИИ в России могут стать вне закона.

根据《Physical Review Letters》期刊上的一项研究，英国 Quantinuum 和荷兰 QuSoft 的研究团队开发出一种量子算法，能比任何经典算法更高效的解决补集抽样（complement sampling）任务，证明量子算法在样本复杂度上具有可证明和可验证的量子优势。想象下有一个巨大盒子，里面装着有编号的球，有人秘密挑选出一半的球组成集合 S，你只能从 S 里取球查看编号，判断哪些编号不在 S 内。经典算法是抽取大量球的样本，才可能有信心判断一个不在 S 中的编号。但量子算法不需要抽取多少样本，你从 S 中抽取的不是一个球，而是一个叠加态的“波球”，通过类似翻转的操作，将 S 的波球翻转到非 S，然后进行测量，就得到了一个不在 S 中的编号。量子算法的抽样次数要比经典算法少得多。