Building security that lasts: Microsoft’s journey towards durability at scale ​​

In this blog you will hear directly from Microsoft’s Deputy Chief Information Security Officer CISO for Azure and operating systems, Mark Russinovich, about how Microsoft operationalized security durability at scale. This blog is part of an ongoing series where our Deputy CISOs share their though...

WhatsApp Adds AI-Powered Message Summaries for Faster Chat Previews

Popular messaging platform WhatsApp has added a new artificial intelligence AI-powered feature that leverages its in-house solution Meta AI to summarize unread messages in chats. The feature, called Message Summaries, is currently rolling out in the English language to users in the United States,...

Malicious code in secure-r007k17 (npm)

The package communicates with a domain associated with malicious activity. --- -= Per source details. Do not edit below this line.=- Source: ghsa-malware bbd04d692c36ca50563b6dbb1de8053730817b3f74b9e23389dd90505621062a Any computer that has this package installed or running should be considered...

undertow: Server identity in https connection is not checked by the undertow client

A flaw was found in undertow. The undertow client is not checking the server identity the server certificate presents in HTTPS connections. This is a compulsory step that should at least be performed by default in HTTPS and in http/2...

wildfly-elytron: possible timing attacks via use of unsafe comparator

A flaw was found in Wildfly-elytron. Wildfly-elytron uses java.util.Arrays.equals in several places, which is unsafe and vulnerable to timing attacks. To compare values securely, use java.security.MessageDigest.isEqual instead. This flaw allows an attacker to access secure information or...

The vulnerability of the Udisks component in the Linux-PAM authentication module allows a hacker to elevate their privileges to the level of root.

The vulnerability of the Udisks component in the Linux-PAM authentication module is related to configuration errors. Exploiting this vulnerability could allow an attacker to elevate their privileges to the root level using the SSH protocol...

UBUNTU-CVE-2025-5372

A flaw was found in libssh versions built with OpenSSL versions older than 3.0, specifically in the sshkdf function responsible for key derivation. Due to inconsistent interpretation of return values where OpenSSL uses 0 to indicate failure and libssh uses 0 for success—the function may mistakenl...

UBUNTU-CVE-2025-4877

There's a vulnerability in the libssh package where when a libssh consumer passes in an unexpectedly large input buffer to sshgetfingerprinthash function. In such cases the bintobase64 function can experience an integer overflow leading to a memory under allocation, when that happens it's possibl...

CVE-2025-6434

The exception page for the HTTPS-Only feature, displayed when a website is opened via HTTP, lacked an anti-clickjacking delay, potentially allowing an attacker to trick a user into granting an exception and loading a webpage over HTTP. This vulnerability affects Firefox 140 and Thunderbird 140...

CVE-2025-6433

If a user visited a webpage with an invalid TLS certificate, and granted an exception, the webpage was able to provide a WebAuthn challenge that the user would be prompted to complete. This is in violation of the WebAuthN spec which requires "a secure transport established without errors". This...

CVE-2025-6433

If a user visited a webpage with an invalid TLS certificate, and granted an exception, the webpage was able to provide a WebAuthn challenge that the user would be prompted to complete. This is in violation of the WebAuthN spec which requires "a secure transport established without errors". This...

CVE-2025-6433

If a user visited a webpage with an invalid TLS certificate, and granted an exception, the webpage was able to provide a WebAuthn challenge that the user would be prompted to complete. This is in violation of the WebAuthN spec which requires "a secure transport established without errors". This...

New Guidance Released for Reducing Memory-Related Vulnerabilities

Today, CISA, in partnership with the National Security Agency NSA, released a joint guide on reducing memory-related vulnerabilities in modern software development. Memory safety vulnerabilities pose serious risks to national security and critical infrastructure. Adopting memory safe languages MS...

K000152032: GRUB2 vulnerability CVE-2025-0678

Security Advisory Description A flaw was found in grub2. When reading data from a squash4 filesystem, grub's squash4 fs module uses user-controlled parameters from the filesystem geometry to determine the internal buffer size, however, it improperly checks for integer overflows. A maliciously...

Secure Multi-Key Homomorphic Encryption with Application to Privacy-Preserving Federated Learning

Whitepaper called Secure Multi-Key Homomorphic Encryption With Application To Privacy-Preserving Federated Learning...

Design High-Confidence Computers Using Trusted Instructional Set Architecture and Emulators

High-confidence computing relies on trusted instructional set architecture, sealed kernels, and secure operating systems. Cloud computing depends on trusted systems for virtualization tasks. Branch predictions and pipelines are essential in improving performance of a CPU/GPU. But Spectre and...

Technical Evaluation of a Disruptive Approach in Homomorphic AI

We present a technical evaluation of a new, disruptive cryptographic approach to data security, known as HbHAI Hash-based Homomorphic Artificial Intelligence. HbHAI is based on a novel class of key-dependent hash functions that naturally preserve most similarity properties, most AI algorithms rel...

SecONNds: Secure Outsourced Neural Network Inference on ImageNet

The widespread adoption of outsourced neural network inference presents significant privacy challenges, as sensitive user data is processed on untrusted remote servers. Secure inference offers a privacy-preserving solution, but existing frameworks suffer from high computational overhead and...

Differential Privacy in Machine Learning: from Symbolic AI to LLMs

Machine learning models should not reveal particular information that is not otherwise accessible. Differential privacy provides a formal framework to mitigate privacy risks by ensuring that the inclusion or exclusion of any single data point does not significantly alter the output of an algorith...

DinoCompanion: an Attachment-Theory Informed Multimodal Robot for Emotionally Responsive Child-AI Interaction

Children's emotional development fundamentally relies on secure attachment relationships, yet current AI companions lack the theoretical foundation to provide developmentally appropriate emotional support. We introduce DinoCompanion, the first attachment-theory-grounded multimodal robot for...