TECHNOLOGY
CYBERSHIELD AI:
INTELLIGENT CYBERSECURITY TRAINING PLATFORM
CyberShield AI is developed by AI Techplus Pty Ltd. It aims to address the growing global threats of cyber fraud and security risks. Using an AI-powered personalized learning system, the platform enables users to master fraud detection, secure online transactions, and cyber threat prevention.
- Smart Fraud Detection:
Detects Credit Card Fraud, Cryptocurrency Scams, Phishing, and Social Engineering Attacks.
- AI-Powered Adaptive Learning:
Personalized learning paths with real-time difficulty adjustment to enhance security awareness.
- Interactive Defense Simulations:
3D attack drills, real-world case studies, and gamified learning experiences.
- Data-Driven Optimization:
Machine learning continuously improves training models, increasing detection accuracy.
Learning Tracks
- Online Fraud Prevention:
From basic pattern recognition to advanced fraud prevention strategies, with progressive training levels.
- Phishing & Social Engineering Defense:
Email analysis, malicious link detection, attack prevention, and response strategies.
REAL-TIME HEART RATE
DETECTION
we used the camera to record 5-10s video and extract video frames to reach 60FPS. Then we utilized Yolov5 and OpenCV to extract the R channel value in RGB in each video frame to detect the colour in real-time, and then used artificial intelligence algorithms to calculate the human heart rate.
Elite Bulter leverages cutting-edge advancements in camera detecton, AI deep learning algorithms, and video action behaviour analysis technologies to meet your needs.
Target Audience
- Corporate Employees:
Enhance company-wide cybersecurity awareness and reduce business risks.
- Developers:
Gain in-depth knowledge of cybersecurity technologies and improve professional skills.
- General Users:
Learn essential fraud prevention skills to protect personal information.
REAL-TIME CAMERA ANALYSIS AND HUMAN BEHAVIOR DETECTION
- OpenPose:
A real-time multi-person pose estimation system basedon deep learning that can detect key points and poses of the human body.
- Action Recognition Models:
A series of deep learning models for action recognition tasks that can recognize and classify various human behaviors.
- PyCoral Action Recognition:
An action recognition model implemented using Google's Coral accelerator that can achieve realtime behavior detection on edge devices.
- DeepLabCut:
An open source toolkit for pose estimation and behavior detection that can be used to study areas such as animal behavior and human motion analysis.
- DensePose:
Facebook's open source human pose and dense pose estimation model that can perform more detailed detection and analysis of human poses and postures.
- By deploying on Alibaba Cloud servers, we can achieve a speed of 30FPS for video action behaviour analysis, and the current validation set accuracy is 92%. With more real data and more training time, our accuracy can reach 99%.
GARBAGE IDENTIFICATION
By using GoogleLeNet, we train the garbage classification network based on a small training data set. We upload the validation model and checkpoints to the Alibaba Cloud platform server so that users can take pictures and upload daily garbage. Our garbage classification algorithm model can complete the upload and classification of the picture within 0.1 seconds, and then transmit the classification detection results back to the user's mobile phone. Currently, we support more than 50 types of daily garbage classification detection. If we have more time, a larger data set and financial support, we can improve GoogleLeNet to support more than 500 types of daily garbage classification detection, and it will be faster and more convenient.
