Unix News

I n recent months, CVE-2024-0132 has emerged as one of the most critical vulnerabilities affecting AI systems, particularly those hosted on cloud environments such as Amazon Web Services (AWS). This high-severity flaw, found within NVIDIA's Container Toolkit, opens the door for attackers to gain full control over a host system by escaping from the container environment. The vulnerability’s potential to wreak havoc on AI workloads, especially when considering the growing use of large language models (LLMs), underscores its importance. As cloud-based infrastructure, such as AWS, becomes the backbone for AI development, the CVE-2024-0132 vulnerability highlights the increasing need for a deep understanding of security best practices for cloud and AI systems. read more..

F ine-tuning Large Language Models (LLMs) has become a crucial step in leveraging the power of pre-trained models for specific applications. This article provides a comprehensive guide on how to fine-tune LLMs using your own data, covering everything from prerequisites to deployment. By the end of this article, you will understand the steps involved in adapting LLMs to meet your unique requirements, enhancing their performance on specialized tasks. read more..

  W indows 10 shell items are metadata files that hold details about various objects in the Windows operating system, including shortcuts, files, and folders. These items are invaluable for forensic investigations because they provide insights into the location and usage of these objects. To perform shell item forensics on Windows 10, you can use forensic tools such as Autopsy, EnCase, or Belkasoft Evidence Center, which are capable of extracting and analyzing shell item metadata. Additionally, manual analysis of shell items is possible using the Windows Shellbags parser, a tool that extracts and interprets the binary data stored in shell item files.. read more...

  A memory leak is a faulty condition where a program fails to free up memory it no longer needs. If left unaddressed, memory leaks result in ever-increasing memory usage, which in turn can lead to degraded performance, system instability, and application crashes. Most modern programming languages include a built-in mechanism to protect against this problem, with garbage collection being the most common. Go has a garbage collector (GC) that does a very good job of managing memory. Garbage collectors such as the Go GC automatically track down memory that is no longer used and return it back to the system. read more...  

  Many Companies Hold Vast Data but Are Unprepared for LLM Fine-Tuning: How to Solve It and What to Do About It In today’s data-driven world, companies across various industries generate and store vast amounts of data. From customer interactions and sales transactions to sensor readings and user-generated content, organizations are sitting on treasure troves of information. However, when it comes to leveraging this data for fine-tuning large language models (LLMs), many companies find themselves unprepared. The growing need for AI-powered solutions requires adapting these models to specific organizational needs—a task that demands both the right infrastructure and expertise. The Challenge: Vast Data, But Lacking Readiness for LLM Fine-Tuning Large language models, such as OpenAI’s GPT or Google’s Bert, have revolutionized industries by providing AI capabilities for natural language understanding, generation, and analysis. However, these models are typically pre-trained on generaliz...

  T he Compute and Performance Engineering teams at Netflix regularly investigate performance issues in our multi-tenant environment. The first step is determining whether the problem originates from the application or the underlying infrastructure. One issue that often complicates this process is the "noisy neighbor" problem. On Titus, our multi-tenant compute platform, a "noisy neighbor" refers to a container or system service that heavily utilizes the server's resources, causing performance degradation in adjacent containers. We usually focus on CPU utilization because it is our workloads’ most frequent source of noisy neighbor issues, read more 

I ncident response and digital forensics play crucial roles in understanding, mitigating, and preventing security events. However, a common pitfall that can undermine even the most sophisticated investigative efforts is the practice of “cherry picking” – selectively choosing evidence that supports a predetermined conclusion while ignoring contradictory information. Whether you’re a seasoned cybersecurity professional or new to the field, understanding the dangers of cherry picking is crucial for conducting thorough and accurate investigations. Let’s dive in and explore why a holistic approach to evidence gathering and analysis is essential in today’s complex threat landscape, read more...

I n today's rapidly evolving cyber threat landscape, organizations face increasingly sophisticated attacks targeting their applications. Understanding these threats and the technologies designed to combat them is crucial. This article delves into the mechanics of a common application attack, using the infamous Log4Shell vulnerability as an example, and demonstrates how Application Detection and Response (ADR) technology effectively safeguards against such zero-day threats.. read more...

  osquery is an open-source security tool that takes an operating system and turns it into one giant database, with tables that you can query using SQL-like statements. With these queries, you can monitor file integrity, check on the status and configuration of the firewall, perform security audits of the target server, and more. It’s a cross-platform application with support for recent versions of macOS, Windows 10, CentOS, and Ubuntu. It’s offically described as an “SQL-powered operating system instrumentation, monitoring, and analytics” framework, and originated from Facebook. read more about it here ..

If you’re running into memory issues because your NumPy arrays are too large, one of the basic approaches to reducing memory usage is compression. By changing how you represent your data, you can reduce memory usage and shrink your array’s footprint—often without changing the bulk of your code. In this article we’ll cover:     * Reducing memory usage via smaller dtypes.     * Sparse arrays.     * Some situations where these solutions won’t work.

    Uv is the "pip but blazingly fast™️ because it's written in rust" and is developed by the same folks that built ruff. It is designed as a drop-in replacement for pip and pip-tools for package management. uv supports everything you'd expect from a modern Python packaging tool: editable installs, Git dependencies, URL dependencies, local dependencies, constraint files, source distributions, custom indexes, and more, all designed around drop-in compatibility with your existing tools. uv's virtual environments are standards-compliant and work interchangeably with other tools — there's no lock-in or customization required. It supports Linux, Windows, and macOS, and has been tested at-scale against the public PyPI index. Read more...

Testing your code brings a wide variety of benefits. It increases your confidence that the code behaves as you expect and ensures that changes to your code won’t cause regressions. Writing and maintaining tests is hard work, so you should leverage all the tools at your disposal to make it as painless as possible. pytest is one of the best tools that you can use to boost your testing productivity. read more...

In this tutorial , you’ll learn: What the differences are between PyTorch and TensorFlow What tools and resources are available for each How to choose the best option for your specific use case You’ll start by taking a close look at both platforms, beginning with the slightly older TensorFlow. Then, you’ll explore PyTorch and some considerations to help you determine which choice is best for your project. Let’s get started!

  This repository contains cutting-edge open-source security tools (OST) that will help you during adversary simulation and as information intended for threat hunter can make detection and prevention control easier. The list of tools below that could be potentially misused by threat actors such as APT and Human-Operated Ransomware (HumOR). If you want to contribute to this list send me a pull request.

This blog post delves into the class of vulnerability known as “Prompt Leaking” and its subsequent exploitation through “Prompt Injection,” which, during an LLM pentest engagement, allowed the unauthorized execution of system commands via Python code injection. In a detailed case study, we will explore the mechanics of these vulnerabilities, their implications, and the methodology used to exploit them.

    eBPF is the new standard to program Linux kernel capabilities in a safe and efficient manner without requiring to change kernel source code or loading kernel modules. It has enabled a new generation of high performance tooling to be developed covering networking, security, and observability use cases. The best way to learn about eBPF is to read the book “What is eBPF” by Liz Rice . And the best way to have your first experience with eBPF programming is to walk through this lab, which takes the opensnoop example out of the book and teaches you to handle an eBPF tool, watch it loading its components and even add your own tracing into the source eBPF code. visit this site and get your hands dirty  ;)

 In the domain of digital forensics, volatile data assumes a paramount role, characterized by its ephemeral nature. Analogous to fleeting whispers in a bustling city, volatile data in Linux systems resides transiently within the Random Access Memory (RAM), encapsulating critical system configurations, active network connections, running processes, and traces of user activities. Once a Linux machine powers down, this ephemeral reservoir of information dissipates swiftly, rendering it irretrievable. read more....

Google DeepMind and Isomorphic Labs today unveiled AlphaFold 3 , a new artificial intelligence model that could greatly accelerate the development of new drugs and treatments. The research published today in Nature reveals AlphaFold 3’s unparalleled ability to predict the intricate structures and interactions of life’s essential molecules, including proteins, DNA and RNA among others. read more...

  This project is about how to systematically persuade LLMs to jailbreak them. The well-known "Grandma Exploit" example is also using emotional appeal, a persuasion technique, for jailbreak! What did we introduce? A taxonomy with 40 persuasion techniques to help you be more persuasive! What did we find? By iteratively applying diffrent persuasion techniques in our taxonomy, we successfully jailbreak advanced aligned LLMs, including Llama 2-7b Chat, GPT-3.5, and GPT-4 — achieving an astonishing 92% attack success rate, notably without any specified optimization . ...