Month: June 2023

As the artificial intelligence (AI) landscape continues to rapidly evolve, new risks and vulnerabilities emerge. Businesses positioned to leverage large language models (LLMs) to enhance and automate their processes must be careful about the degree of autonomy and access privileges they confer to LLM-powered AI solutions, wherein lies a new frontier of cybersecurity challenges. 

In this article, we take a closer look at prompt hacking (or prompt injection), a manipulation technique through which users may potentially access sensitive information by tailoring the initial prompt given to a language model. In the context of production systems that house a wealth of sensitive data in databases, prompt hacking poses a significant threat to data privacy and security from malicious actors. A successful prompt hacking attack against these resources could enable unauthorized reading or writing of data, leading to breaches, corruption, or even cascading system failures.

This article aims the following audiences:

• Tech workers
• Students
• Product managers
• Marketing team
• Sales team

There are no prerequisites to reading this article. I assume that you have used a web browser to navigate across the internet. 

I think everyone at least once used the explaincommand or at least heard about it. This command demonstrates the query execution plan, but how exactly the DBMS arrives at it remains a mystery. And how does the DBMS understand that the selected query is optimal? Does it check all possible options?

In this article, I will try to give a little idea of how query optimizers work from a theoretical point of view.

As our world becomes increasingly automated, the role of artificial intelligence (AI) in various domains has grown exponentially. While AI brings numerous benefits and advancements, it also introduces new threats and challenges in cybersecurity. This article will delve into the unseen challenges that AI threats pose, exploring the potential risks and vulnerabilities associated with this technology.

Understanding AI Threats

AI has revolutionized healthcare, finance, transportation, and more industries. However, its rapid integration into critical systems and decision-making processes allows malicious actors to exploit vulnerabilities. As a cybersecurity professional with 20 years of experience, I have witnessed firsthand the evolving landscape of cyber threats, and AI presents unique challenges that demand our attention.

In today's rapidly evolving cyber threat landscape, traditional reactive approaches to cybersecurity are no longer sufficient. As attackers become more sophisticated, organizations must adopt proactive defense strategies to stay one step ahead. This article will delve into the changing face of threat hunting in 2023 and explore how cybersecurity professionals can master aggressive defense to combat emerging threats effectively.

The Evolution of Threat Hunting

Threat hunting has evolved from a reactive approach to a proactive one. Traditionally, organizations rely on security tools that detect known threats or anomalies after an attack. However, aggressive defense has become essential with the increasing complexity and frequency of attacks. Threat hunting now involves actively searching for signs of compromise within an organization's network and systems before an attack occurs.

Fluent Bit is a popular open-source log processor and forwarder that allows you to collect data from different sources, filter, and transform it before forwarding it to different destinations. In some cases, the data collected may contain sensitive information like passwords, credit card numbers, social security numbers, and other personally identifiable information (PII). To protect such information, you need to mask or obfuscate it before forwarding it to the destination. In this document, we will discuss how to mask sensitive information using Fluent Bit.

The goal of this guide is to convert structured logs that contain PII information like (mobile numbers, identity information, names, etc.)
{"timestamp":"2023-06-05T17:04:33.505+05:30","requestURI":"/api/user","message":"Sending SMS to mobileNumber=1234512345 registered on aadhaarNumber=1234512345"}
to a format where this information is masked.
{"@timestamp":"2023-06-05T17:04:33.505+05:30","requestURI":"/api/user","message":"Sending SMS to mobileNumber=******** registered on aadhaarNumber=********"}

When I discovered OpenAI on the 23rd of December 2022, I became obsessed. I hadn't had this much fun coding since I started 40 years ago. After playing with "fine-tuning" for a month and miserably failing, I found a YouTube video by Dave Shapiro that explained how to create a Q&A chatbot using OpenAI's "embeddings" API.

Dave has since removed the video, but it explains how to use OpenAI's embeddings API and combine it with their chat/completion API to create a Q&A chatbot that knows "everything" about the problem domain. Tage's reaction best sums up my findings, as he told me one day.

Understanding Deadlocks in Databases

In the context of databases, deadlocks are situations where two or more transactions are unable to proceed because each is waiting for the other to release a resource. This can be likened to a circular chain of transactions, where each transaction is waiting for the next in the chain to release a resource.

Here’s a visual representation of a deadlock scenario:

Changing an image/video background has always been a hassle, whereby the most tricky part is extracting the element other than the background.

Traditionally, it requires us to use a PC image-editing program that allows us to select the element, add a mask, replace the canvas, and more. If the element has an extremely uneven border, then the whole process can be very time-consuming.