Maritime 4.0: Innovation Driven by AI, Data, and Cyber Security

In February 2024, OpenAI introduced Sora , an advanced text-to-video AI capable of generating highly realistic videos up to one minute long based on textual descriptions. This revolutionary model has raised cybersecurity concerns , particularly in the realm of deepfake attacks, misinformation, and AI-driven cyber threats . ( dbr.donga.com ) 1. Deepfake Cyber Threats & Social Engineering Risks Sora’s ability to create hyper-realistic videos has significantly lowered the barrier for cybercriminals to craft deceptive content . This includes: Impersonation Attacks: Hackers could generate deepfake videos of executives, politicians, or security personnel to manipulate decision-making processes. Spear Phishing & Business Email Compromise (BEC): Attackers could integrate AI-generated videos into phishing campaigns, making social engineering tactics more convincing . Disinformation & Psychological Warfare: Nation-state actors may use Sora to create manipulated news reports or...

In recent years,  "NL2SQL"  has gained attention as a groundbreaking technology that enhances interaction with databases. NL2SQL stands for  Natural Language to SQL , a  natural language processing (NLP) technique that converts human language into SQL queries . This technology is emerging as a  powerful cybersecurity tool , particularly in  cyber threat detection and maritime hacking defense .   1. Enhancing Cybersecurity Monitoring with NL2SQL NL2SQL streamlines security monitoring and threat detection by enabling intuitive queries in natural language. 🔹 Security Operation Centers (SOC) can instantly identify threats with queries like: "Show me all abnormal login attempts in the last 24 hours." 🔹 SQL Conversion Example: 복사 편집 SELECT * FROM login_attempts WHERE timestamp >= NOW() - INTERVAL 24 HOUR AND failed_attempts > 5 ORDER BY timestamp DESC ; This allows security teams to detect potential hacking attempts quickly wi...

I have resumed engineering and modeling tasks , which I had temporarily set aside to focus on project management and new business planning . For a project that involves optimizing (tuning) individual models for image and text processing , as well as integrating multiple models —a skill I honed during my Ph.D. program—I now require a complex and well-structured work environment . To ensure optimal efficiency under the given constraints , I plan to set up, refine, and document the work environment for future reference. "If I had eight hours to chop down a tree, I would spend six hours sharpening the axe." – Abraham Lincoln (1861–1865), the 16th President of the United States Review of Work Environment 1. Managing Work on an Unstable Server & Data Backup ✅ Backup & Record Management: Using SFTP to store and synchronize data in a structured manner. ✅ GitHub Usage: Limited to baseline model development and version control for key components. ✅ Code & Data Managem...

For decades, maritime vessels operated with limited and slow satellite communications , making them less attractive targets for cyberattacks . However, with the introduction of Starlink , ships now have access to high-speed, low-latency internet , transforming them into fully connected digital platforms . While this advancement brings enormous benefits in operational efficiency, it also exposes vessels to unprecedented cybersecurity risks .  SpaceX Debuts Maritime Offering for Starlink As the shipping industry embraces Starlink-powered connectivity , investing in cybersecurity has become a necessity, not an option . This article explores the growing cyber threats faced by ships due to Starlink and how businesses should strategically invest in security measures to mitigate these risks. 🌊 1. How Starlink is Changing the Cybersecurity Landscape for Ships Traditionally, ships relied on expensive and slow VSAT (Very Small Aperture Terminal) or Inmarsat satellite connections , which...

I would like to organize and document the learning curriculum and key research papers of interest for the Recommender System course that I am taking in the final semester of my Ph.D. program. Goals of this Course Collaborative Filtering (CF) A recommendation approach based on user behavior . Neighborhood-based Collaborative Filtering (CF) A method that recommends items based on similarities between users or items . User-based Collaborative Filtering (User-based CF) Recommends items based on similarities between users . Users with similar preferences or behaviors receive similar recommendations. Item-based Collaborative Filtering (Item-based CF) Recommends items based on similarities between items . Items that have been interacted with in a similar way are recommended together

 I have come to As the first step, I am recording my personal conclusions on the issues I have contemplated over the past month. [Relational Aspects] I do not strive to become someone else in order to be remembered as a good person in the organization. I am who I am. Work-related matters should be discussed in meetings where all relevant team members (or stakeholders) are present. Even if someone feels uncomfortable or is being held accountable, I will not avoid the situation. Instead, I will share the facts accurately and discuss solutions together to foster future growth and stronger relationships. When I recognize actions that go against basic common sense, I will focus on identifying the fundamental issue rather than getting caught up in others’ emotions. I do not indiscriminately offer consideration, comfort, or reprimands to others. I focus solely on facts and do not convey my emotions in judgment. The driving force of my 30s— “An eye for an eye, a tooth for a tooth” ...

Shifting Perceptions and Expectations of AI from Companies The way companies perceive and expect from AI is evolving. Until now, aside from cutting-edge research and product development , the role of AI engineers and data scientists in enterprises can generally be categorized as follows: Image Processing : Companies have widely utilized CNN-based models for object recognition and comparison (performance improvement). To overcome data scarcity, techniques like few-shot learning have been applied in areas such as product search, similar image/product recommendation, and design generation . Text Processing : Enterprises have leveraged various Transformer-based models trained on proprietary datasets for customer intent recognition and sentiment analysis . Data-Driven Decision Support : Companies often assume they possess sufficient data, but in many cases, they lack the necessary datasets. AI teams analyze available data, form hypotheses , and generate insightful reports to support an...

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks EfficientNet is a highly influential paper that has gained significant attention in the field of image classification due to its outstanding performance . For projects requiring extensive training time or computational resources , EfficientNet serves as a valuable approach to enhancing ConvNet performance . It provides an efficient and scalable method for training convolutional neural networks while optimizing accuracy and computational cost, making it highly applicable for real-world AI deployment . 🔗 Research Paper: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks 📌 Key Resources & Reviews 📖 Paper Review Summaries: Bellzero’s Review Laonple Blog Review 💻 Source Code (PyTorch Implementation): GitHub: EfficientNet-PyTorch  

Installing Required Libraries to Convert SVG Files to PNG The goal was to apply NBP OCR to SVG images and record the output as JSON (text, coordinates) . To achieve this, the first step was to convert the ".svg" file into ".png" . Initially, I thought that simply installing cairo via pip would suffice, but I soon realized that additional steps were necessary . To avoid wasting time solving issues when I attempted it again later, I decided to document the process and share it . If you need code references or encounter any setup issues, please feel free to leave a message here: https://github.com/shipjobs/HAND2TEXT/issues . Environment Language: Python 3.8.3 64bit Operating System: Windows 10 Development Environment: Visual Studio Code To convert an ".svg" file to ".png" , if you import the libraries as shown below, you will naturally encounter a reference error : import  cairo  from  svglib.svglib  import  svg2rlg  import  cairosvg For those...

CLOVA OCR (optical character reader) Service review 문서를 인식하고, 사용자가 지정한 영역의 텍스트와 데이터를 정확하게 추출 CLOVA OCR (광학문자인식) 을 한번 사용해 본다면, NCP가 제공 중에 있는 AI Service 들 에 대하여 보다 쉬운 접근이 가능해질 거라 생각 하게 되어 review 를 해보고자 합니다. [접근] Products & service : https://console.ncloud.com/dashboard OCR Service 경로 : Classic / CLOVA OCR / Domain [이용 방식] 서비스 타입 General / Template / Document 선택에 따라 Text OCR / 템플릿 빌더 / Document 버튼이 노출되는 형식으로 서비스를 설정함  Text OCR (텍스트만 추출)  과 Template 빌더 형태 (판독 영역 직접 지정을 통해 인식 값 추출 후 테스트 및 결과 전송이 가능) 는 서비스 타입에 따라 아래의 2가지 방식이 있으며 1. General OCR : 우리가 일반적으로 생각하는 png, jpg이미지 혹은 pdf 에 존재하는 text 들을 모두 읽어 오고자 하는 방식 2. Template OCR : 운전 면허증, 신용 카드, 주민 등록 등본 이미지 등 이미지내 정해진 특정 영역을 기준으로 text 들을 읽어 오는 방식 Document 방식은 머신러닝 기반으로 문서의 의미적 구조를 이해하는 특화 모델 엔진을 탑재하여 입력 정보(key-value)를 자동 추출하는 방식 인식 모델로는 미리 정해진 사업자 등록증, 신용카드,영수증, 신분증, 명함이 제공 되며 이를 선택할 수 있게 되어 있습니다. [이러한 Type 별 서비스를 이용하는 방법 또한 2가지로 구분 될 수 있습니다.] 1. NCP OCR 사이트에 접속해서 제공되는 UI 화면으로 접근 하는 방식으로 원하는 이미지 파일을 drag and drop으로 등록하고 텍...