Tech Unplugged

This podcast analyzes the susceptibility of modern language models to various attack techniques , revealing vulnerabilities at both the textual and architectural levels despite existing safeguards. The author emphasizes the models' inherent trust and literal command execution as key exploitable traits. To mitigate these risks, the text proposes several short-term recommendations for developers and...

This podcast introduces Portkey , an AI gateway and observability platform designed to streamline the development and deployment of AI applications. It highlights features such as managing API access to multiple LLM providers , offering tools for monitoring performance, costs, and usage , and providing capabilities for prompt management, caching, fallbacks, and security guardrails . Additionally,...

This podcast explore the burgeoning field of LLM firewalls as a critical security measure for applications utilizing large language models. These sources highlight the unique risks associated with LLMs , such as prompt injection, data leakage, and model abuse, which traditional firewalls are ill-equipped to handle due to the integrated nature of data and operations within LLMs . Several companies,...

Engineers addressed the problem of inaccurately attributing network flow logs to specific workloads in their cloud environment. Their initial system, relying on IP address change events, suffered from misattribution due to delays in event propagation. To resolve this, they implemented a new method using eBPF on workload instances to directly associate local IP addresses with workload identities, s...

This podcast provides in-depth explanations and hands-on exercises across areas such as identity and access management with Azure AD, network security implementation using firewalls and WAF, securing compute and storage services including containers and databases, and managing security operations with Azure Monitor and Sentinel. Both resources emphasize practical application and understanding of A...

intersection of open-source red teaming tools and malicious cyber activities conducted by cybercriminals and nation-state actors. It highlights the dual-use nature of these tools , originally intended for security testing, and their increasing adoption in sophisticated attacks, including supply chain compromises. The paper discusses methodologies for managing open-source threats , including AI-bas...

The podcasts collectively explore the Windows API from various perspectives, including its fundamental role in Windows programming, its use in red teaming and cybersecurity , and considerations for testing applications that utilize it, specifically large language models. C++ examples for red team operations using Windows APIs, to a guide on leveraging these APIs in C# with P/Invoke , and discussio...

CVE-2025–29927 , a critical security flaw in the Next.js web framework. The author, coffinxp , details how this vulnerability allows attackers to bypass middleware authorization , potentially leading to unauthorized access to protected resources. The article clarifies the purpose of Next.js middleware and how the specific flaw in its request handling enables this bypass. Furthermore, it suggests t...

This podcasts extensively cover industrial cybersecurity , focusing on monitoring, threat hunting, vulnerability assessments, and incident response within industrial control systems (ICS) and operational technology (OT) environments. A significant portion details the architecture and security of ICS networks , including the importance of network segmentation and the industrial demilitarized zone (...

Vector databases are introduced as a solution to the limitations of traditional databases when handling unstructured data by representing it as vector embeddings , which are numerical arrays capturing semantic meaning. These databases enable similarity searches based on conceptual relationships rather than exact matches. Embedding models , trained on vast datasets, generate these vector embeddings...

This podcast explains the fundamental concepts of AI agents and their ability to automate tasks by utilizing various tools. This podcast demonstrates the practical application of these concepts through four distinct no-code agent-building projects using platforms like Relevance AI, n8n, Voiceflow, and customized own platform, Agentive. These builds cover use cases such as a sales co-pilot, automat...

Several sources discuss software vulnerabilities across different programming languages. One source highlights a specific vulnerability, CVE-2024-27322, in the R programming language . Another examines the potential of memory-safe languages like Rust, Go, and Python to mitigate a significant portion of security bugs prevalent in languages such as C and C++. Additionally, one article ranks the top...

Core concepts like pods, deployments, services, and nodes , while also covering more advanced topics such as ConfigMaps, Secrets, Helm, and cluster management . Rudi Martinsen's blog post specifically examines Ingress , detailing its function in routing external traffic and the necessity of Ingress controllers . practical kubectl commands and example configurations .Kubernetes topics, including TL...

Agent Reasoning Hijacking affecting LLM agents that use chain-of-thought reasoning and external tools. This flaw allows attackers to inject adversarial strings that manipulate the agent's thinking process, leading it to perform unintended malicious actions like data theft or unauthorized access. The sources detail how this attack works, its potential impact on various LLM models and real-world app...

This podcast offer a comprehensive overview of MongoDB, a NoSQL database, highlighting its flexible data handling, scalability through sharding, and features like implicit collection creation and document manipulation using methods like insertOne() and deleteOne(). Security aspects, including data encryption at rest and the importance of regular backups, are discussed alongside its support for uns...

This podcast AI-specific threats like data poisoning and prompt injection , differentiating them from traditional security concerns due to AI's non-deterministic nature and evolving vulnerabilities . The sources also discuss the application of AI to enhance security practices, such as threat detection, incident response, and automation , while highlighting the necessity of securing AI systems them...

NVIDIA Cosmos is a platform detailed in a research paper and related NVIDIA resources, aiming to advance Physical AI for robots and autonomous vehicles. It provides pre-trained World Foundation Models (WFMs) capable of generating future video frames from various inputs like images and text, addressing the challenge of limited real-world training data by enabling the creation of numerous simulated...

Security architecture reviews for integrating SaaS applications . It outlines a structured process , including defining scope, assessing provider security, and considering data, identity, API, and compliance aspects. Furthermore, it features questionnaires and checklists to guide the review process, ensuring comprehensive evaluation of potential risks and adherence to security best practices and r...

Microsoft SQL Server as a relational database management system designed for storing and retrieving data for various software applications. It outlines the client-server architecture of SQL Server, highlighting three core elements: the Protocol Layer for connectivity, the Relational Engine for query processing, and the Storage Engine for data management. The text further explains the components wi...

PostgreSQL , an open-source relational database system. It introduces fundamental concepts , such as installation, database creation, table design with data types and constraints, and basic data manipulation through SQL queries. The guide also touches upon more advanced topics like indexing for performance and the importance of database backup and restoration. Through a practical example , the aut...

Structured Query Language (SQL) , a standard language for database interaction. It meticulously details various SQL commands , categorizing them into DDL for structure definition, DML for data manipulation, DCL for access control, TCL for transaction management, and DQL for data retrieval. The document further explains fundamental SQL operations like creating and modifying tables, inserting, updat...

This podcast cover a range of cybersecurity vulnerabilities and attack techniques. One source details the reverse engineering of an Android application leading to a remote code execution exploit. Another explores a novel perspective on Server-Side Request Forgery for account takeover. Cross-Window Forgery, a new class of web attack exploiting HTML ID attributes, is also examined. Additionally, the...

In this podcast, we explore Windows persistence mechanisms used by Red Teams to maintain access in compromised systems. We’ll cover common tactics such as registry modifications, scheduled tasks, WMI persistence, DLL hijacking, and user account manipulation. Learn how adversaries leverage these techniques to evade detection and ensure long-term control. We’ll also discuss real-world examples, dete...

This research investigates the use of autonomous systems driven by Large Language Models (LLMs) for Assumed Breach penetration testing in enterprise networks. The authors developed a novel prototype capable of compromising accounts within a real-life Active Directory testbed. Their evaluation highlights the prototype's strengths and limitations in simulating attacks, using a realistic environment...

This podcast offers a comprehensive overview of Azure App Registrations , a key component for secure application interaction with Azure services. It begins by detailing the various types of Azure application registrations , explaining their characteristics and use cases. The report then thoroughly examines the risks and security concerns associated with improper configuration and management of the...