Image by Kevin Ku, from Unsplash
AI Model Achieves 98.53% Accuracy In Detecting Ransomware On Smart Devices
- Written by Kiara Fabbri Former Tech News Writer
- Fact-Checked by Sarah Frazier Former Content Manager
Scientists developed an AI model detecting ransomware in IoT devices with high accuracy, using deep learning and optimization techniques for cybersecurity.
In a Rush? Here are the Quick Facts!
- The AI model detects ransomware in IoT devices with 98.53% accuracy.
- It uses min-max normalization and dung beetle optimization for better threat detection.
- Multi-head attention and LSTM networks analyze ransomware patterns to predict attacks.
A team of researchers has detailed their findings today in a Scientific Reports paper published by Nature, introducing an advanced AI-powered model designed to detect and prevent ransomware attacks on smart devices.
With the rapid expansion of Internet of Things (IoT) technology in homes, healthcare, and industries, cybersecurity threats have become a growing concern.
Ransomware, one of the most dangerous cyber threats, locks users out of their systems until they pay a ransom. The researchers explained how traditional security measures often fail to detect and prevent these evolving attacks, prompting researchers to explore AI solutions.
Their newly developed model, called Multi-head Attention-Based Recurrent Neural Network with Enhanced Gorilla Troops Optimization (MHARNN-EGTOCRD), significantly improves ransomware detection accuracy using machine learning techniques.
The model first normalizes incoming data using min-max normalization, ensuring efficient processing. It then employs Dung Beetle Optimization (DBO)—inspired by how dung beetles locate food—to filter out unnecessary information, focusing only on the most relevant cybersecurity threats.
At its core, the system utilizes a Multi-head Attention and Long Short-Term Memory (MHA-LSTM) network, an advanced deep learning approach that helps detect complex attack patterns.
By analyzing past ransomware behaviors, the AI can predict and flag potential attacks before they fully execute. Additionally, the system is fine-tuned using Enhanced Gorilla Troops Optimization (EGTO), which optimizes the AI’s settings for maximum efficiency.
In testing, the model achieved an impressive 98.53% accuracy in detecting ransomware, outperforming traditional cybersecurity methods. This high precision suggests that AI could become a powerful tool in the fight against cybercrime, particularly in safeguarding smart devices from sophisticated attacks.
The researchers believe their model could be integrated into existing cybersecurity systems, providing an early warning mechanism for ransomware attacks.
As IoT devices continue to expand in everyday life, strengthening their security is crucial to preventing financial and data losses. By combining nature-inspired optimization techniques with deep learning, this AI model represents a significant step forward in cybersecurity.
Image by JD Lasica, from Wikimedia Commons
Musk’s DOGE Deploys Chatbot After Firing Workers
- Written by Kiara Fabbri Former Tech News Writer
- Fact-Checked by Sarah Frazier Former Content Manager
Elon Musk’s Department of Government Efficiency (DOGE) is rolling out a proprietary chatbot, GSAi, to automate tasks for federal workers amid massive job cuts.
In a Rush? Here are the Quick Facts!
- GSAi helps employees with emails, summaries, talking points, and coding tasks.
- Workers can’t use GSAi for nonpublic or controlled unclassified information.
- Employees describe GSAi as “about as good as an intern” with generic answers.
Around 1,500 employees at the General Services Administration (GSA) now have access to the AI tool, which aims to replace some of the work previously handled by terminated or resigned staff, as first reported by WIRED .
The chatbot’s deployment comes as DOGE accelerates its workforce reductions, which some experts have called “the largest job cut in American history,” as noted by Gizmodo .
GSAi, similar to ChatGPT, is designed to assist with routine administrative tasks. According to an internal memo, reported by WIRED, employees can use it to “draft emails, create talking points, summarize text, [and] write code.”
However, workers are restricted from entering nonpublic or controlled unclassified information, limiting its usefulness. One GSA employee told WIRED that the chatbot is “about as good as an intern” and provides “generic and guessable answers.”
The chatbot had been in development before DOGE’s involvement, with the Treasury and the Department of Health and Human Services previously exploring chatbot interfaces for internal use.
WIRED reports that the Department of Education was also collaborating with GSA on a chatbot for support tasks. However, an employee described these projects as “janky,” and they had not been widely deployed until DOGE hastened their rollout.
In February, DOGE tested GSAi with 150 users before expanding its availability. Employees interact with the chatbot through an interface resembling ChatGPT, with models such as Claude Haiku 3.5, Claude Sonnet 3.5 v2, and Meta LLaMa 3.2 available depending on the task, as reported by WIRED.
“What is the larger strategy here? Is it giving everyone AI and then that legitimizes more layoffs?” an AI expert questioned in WIRED. “That wouldn’t surprise me.”
The move to deploy AI coincides with sweeping reductions in the government’s tech workforce. At a recent town hall, Thomas Shedd, a former Tesla engineer now leading the GSA’s Technology Transformation Services (TTS), announced that the agency’s tech division would shrink by 50% in the coming weeks, said WIRED.
Last week, around 90 technologists were let go. Moving forward, TTS will focus on public-facing projects like Login.gov and Cloud.gov, cutting all other non-mandated work. “We will be a results-oriented and high-performance team,” Shedd stated, reinforcing the administration’s push for AI-driven efficiency, as reported by WIRED.
However, with thousands of workers already dismissed, some question whether AI can truly replace the expertise lost.
