More doctoral student excellence: Congrats to Liangyi Huang whose paper was recently accepted by the Conference on Language and Modeling (#COLM) which will be held at the University of Pennsylvania in October. In the work, conducted under the supervision of Xusheng Xiao, Huang proposes a novel way — powered by Large Language Models — to achieve 90% accuracy in building security-oriented knowledge graphs from Cyber Threat Intelligence articles written in natural language. COLM is an academic venue focused on the study of language modeling, broadly defined, with the goal of creating a community of researchers with expertise in different disciplines, focused on understanding, improving, and critiquing the development of LM technology. Congrats Liangyi! Visit the team’s project website: https://lnkd.in/gJwHtd_S. Learn more about Liangyi: https://lnkd.in/gFN3UufT #SCAI #PhD
School of Computing and Augmented Intelligence’s Post
More Relevant Posts
-
UI/UX Designer || ex AI Research Intern @GBJBuzz | SIH '23 Grand Finalist | IEEE Ideathon '22 Finalist
Hello, LinkedIn Connection! 🌐 📄 I am thrilled to announce the publication of my research paper in the International Journal of Computer Research and Technology IJCRT! 🎉 🔗 Read the full paper here: https://lnkd.in/g67jTWER 🎓Final Year Major Projects are always a significant responsibility, marking the culmination of your college journey where you demonstrate the knowledge gained during your graduation. I, along with my colleagues Tanvi Jikar and Trupti Kularkar, pursued our project "Deepfake Detection Using ResNext and LSTM" under the guidance of Krupali Dhwale ma'am. Excited to share that we have successfully published our research paper in IJCRT! 🚀 🤥In a world where digital content reigns supreme, the rise of deepfakes has become a concerning challenge. These deceptive manipulations of audiovisual material can sow seeds of misinformation, erode privacy, and even pose threats to national security. With technology evolving at a rapid pace, our collective vulnerability to the malicious use of deepfakes necessitates innovative solutions. 🔮In our research paper, we dissected this intricate problem and sought to address it head-on. Leveraging the power of ResNext and LSTM, we devised a robust solution to detect deepfakes effectively. Our approach involves meticulous analysis, utilizing advanced neural networks to discern authentic content from manipulated ones. The ResNext architecture enhances feature extraction, while LSTM's memory capability allows for a nuanced understanding of sequential data, crucial for identifying subtle anomalies in videos. 💡🔒This isn't just about a research paper; it's about contributing to the defense against a growing digital threat. We believe that by developing and refining technologies like ours, we can fortify the digital landscape against the deceptive potential of deepfakes. . #DeepfakeDetection #DigitalSecurity #Innovation #ResNext #LSTM" 🔗Attached is our certificate of accomplishment, a testament to our dedication and hard work. Special thanks to our guide, Krupali Dhawale, ma'am, for her invaluable support. 🙌I am grateful for the opportunity to share our findings and contribute to the field. Looking forward to continued learning and innovation! #ResearchPaper #Collaboration #knowledgesharing
To view or add a comment, sign in
-
-
🎉 Congratulations to Alexander Unnervik (Alex U.) from Idiap Research Institute on the successful defense of his PhD thesis “Performing and Detecting Backdoor Attacks on Face Recognition Algorithms” under the supervision Prof. Jean-Philippe Thiran and Dr Sébastien Marcel. 🔎 Deep learning algorithms are vulnerable to Backdoor Attacks. Targeted networks behave normally except when the attacker takes control of the model output, at will. This thesis shows that open-set face recognition algorithms are prone to these backdoor attacks and proposes two methods to detect the presence of these attacks in pretrained face recognition networks. #epfl_edee
To view or add a comment, sign in
-
-
Cyber Security Professional | SOC Analyst | Threat Hunting | Digital Forensic | Incident Responder | SIEM | LogRhythm | Wazuh | EDR/XDR | Malware Analysis | MITRE ATT&CK | Cyber Kill Chain | DLP
Alhamdulliah. I am excited to share that my first research paper, "DivaCAN: Detecting in-vehicle intrusion attacks on a controller area network using ensemble learning," got accepted in the Computers & Security Journal (Impact Factor = 5.6, CiteScore = 11.1). It was also my Master's thesis, focusing on CAN security and introducing DivaCAN to counter denial-of-service, fuzzy, and impersonation attacks on the CAN bus. DivaCAN uses various classifiers, including deep neural networks and random forest, for robust intrusion attack recognition. I would like to express my sincere gratitude to my esteemed co-authors, Abdul Rehman Javed, Dr. Zafar Iqbal, Muhammad Asim, and Ali Ismail Awad, as well as the anonymous reviewers, the journal Associate Editor, and the journal Editor-in-Chief for their invaluable contributions to this work. You are warmly invited to peruse the paper by following the link below: https://lnkd.in/d_HV3Unq
To view or add a comment, sign in
-
-
We just had an enlightening discussion in our reading group about a recent paper on Prompt-Based Adversarial Attacks, featuring one of our colleagues, JINGFENG ZHANG, as one of the authors! JINGFENG ZHANG is an outstanding researcher and lecturer at the School of Computer Science - University of Auckland. This is a great read that inspired extensive discussions in our lab concerning the implications of their findings and the potential for future work. We're eager to see what comes next from these remarkable authors. Check out their paper here: https://lnkd.in/gfkWKXfa What do you think are the most pressing challenges and opportunities in the realm of adversarial attacks? Share your thoughts below! #MachineLearning #AdversarialAttacks #ArtificialIntelligence #ResearchImpact
To view or add a comment, sign in
-
Trustworthy Machine Learning, Adversarial Robustness, Robust Foundation Models & Reliable Fine-Tuning Strategies
Many thanks for sharing and reading this ICLR2024 paper, that I lead as corresponding author 😀 . This paper shows how we can leverage LLM to craft NLP adversarial examples with minimum notice. Emoji attack is an example :). This paper has provided a principled approach on how to craft such the examples at different levels, such as adding a single character, changing a single word, or rephrasing a sentence. This is project page where you can play the code https://lnkd.in/dcQjhF6A
We just had an enlightening discussion in our reading group about a recent paper on Prompt-Based Adversarial Attacks, featuring one of our colleagues, JINGFENG ZHANG, as one of the authors! JINGFENG ZHANG is an outstanding researcher and lecturer at the School of Computer Science - University of Auckland. This is a great read that inspired extensive discussions in our lab concerning the implications of their findings and the potential for future work. We're eager to see what comes next from these remarkable authors. Check out their paper here: https://lnkd.in/gfkWKXfa What do you think are the most pressing challenges and opportunities in the realm of adversarial attacks? Share your thoughts below! #MachineLearning #AdversarialAttacks #ArtificialIntelligence #ResearchImpact
An LLM can Fool Itself: A Prompt-Based Adversarial Attack
arxiv.org
To view or add a comment, sign in
-
I thought the :) emoji attack was an interesting approach. JINGFENG ZHANG, does it work with other non-emoji based punctuation? *> or #- Yonghua Zhu, thanks for presenting this paper!
We just had an enlightening discussion in our reading group about a recent paper on Prompt-Based Adversarial Attacks, featuring one of our colleagues, JINGFENG ZHANG, as one of the authors! JINGFENG ZHANG is an outstanding researcher and lecturer at the School of Computer Science - University of Auckland. This is a great read that inspired extensive discussions in our lab concerning the implications of their findings and the potential for future work. We're eager to see what comes next from these remarkable authors. Check out their paper here: https://lnkd.in/gfkWKXfa What do you think are the most pressing challenges and opportunities in the realm of adversarial attacks? Share your thoughts below! #MachineLearning #AdversarialAttacks #ArtificialIntelligence #ResearchImpact
An LLM can Fool Itself: A Prompt-Based Adversarial Attack
arxiv.org
To view or add a comment, sign in
-
Applying Machine Learning Techniques in Cryptographic Algorithms 29-May-2023 Excited to share insights from my Bachelor thesis, where I integrated machine learning techniques with cryptography to bolster data security. This project not only deepened my understanding of both fields but also significantly enhanced my research and writing skills into the world of neural networks. Through rigorous analysis of various Deep Learning architectures, I utilized Tensorflow’s Keras to implement the models as well as Google Colab's environment for the computational boost and efficiency. Key points: - Explored vulnerabilities in Simplified Data Encryption Standard (S-DES) and Speck algorithms. - Achieved a high accuracy in deciphering ciphertexts with plaintext-ciphertext pairs. - Attained an average Bit Accuracy Probability (BAP) of over 90% for key prediction for 10 bit keys. - Compared effectiveness on longer key sizes between algorithms. Celebrating the end of this journey, the defence presentation was an opportunity to showcase and discuss my work in front of esteemed professors and experts. It was a moment of pride to receive an A grade for the project. Find the full paper here: https://lnkd.in/dVvTpdVf
To view or add a comment, sign in
-
-
Dear Connections!!! 🎓 I am excited to unveil our Capstone Project Poster: Title: "ML-Integrated Smart Camera 📷 " Our project is reshaping surveillance by harnessing the power of Machine Learning (ML) for efficient and versatile security solutions. Key Features: 1. Efficiency Meet Versatility: - Activates video recording only on motion detection, minimizing unnecessary memory usage. - Advanced human recognition with algorithms like Haar Cascade, face recognition, and background subtraction for movement detection. 2. OpenCV and ML: - Utilizes OpenCV for real-time tasks, image processing, and ML algorithms. - Enhances accuracy with data containing images of known individuals. Related Work: Explores various ML applications, face encoding, face detection, and deep learning for threat detection. Acknowledgments: Thanks to our mentor sandeep chintham and dedicated teammates - spoorthi gangasani, Raju Bayya, Deepika Danda, Donikena Dharani, and konur sandeep. CS&AI -SRU | SR University |Dr.Sheshikala Martha Proudly presented under CS&AI at SR University, contributing to the evolution of security measures. #sru #sruniversity #ml #security #cse
To view or add a comment, sign in
-
-
Fresh Graduate of Informatics at Telkom University | Ex-Research Assistant at BRIN RI | Data Enthusiast
Exciting News! 🎉 My research paper, titled "Hyperparameter Optimization in Convolutional Neural Network for Sentiment Analysis: A BSI Ransomware Perspective," has been successfully published on the IEEE Xplore digital library! This research explores the use of Convolutional Neural Networks (CNNs) for sentiment analysis related to BSI ransomware. The research was presented at the proceedings of the 2024 2nd International Conference on Software Engineering and Information Technology (ICoSEIT). I'm thrilled to share this contribution to the field and hope it inspires future research endeavors. Read the full paper here: https://bit.ly/3xO2R92 #ICoSEIT2024 #Research #SentimentAnalysis #DeepLearning #CNN
To view or add a comment, sign in
-
Looking back at the #EuAIAct and the surpsie of regualtion of #AI for industries, we should also look back into the past. Have a look at the recommendations of the publication "The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation " https://lnkd.in/dhmNPQD6 by Future of Humanity Institute (Oxford University) University of Oxford, Centre for the Study of Existential Risk, University of Cambridge, University of Cambridge, Center for a New American Security (CNAS), Electronic Frontier Foundation (EFF) and OpenAI from February 2018. 1. Policymakers should collaborate closely with technical researchers to investigate, prevent, and mitigate potential malicious uses of AI. 2. Researchers and engineers in artificial intelligence should take the dual-use nature of their work seriously, allowing misuse- related considerations to influence research priorities and norms, and proactively reaching out to relevant actors when harmful applications are foreseeable. 3. Best practices should be identified in research areas with more mature methods for addressing dual-use concerns, such as computer security, and imported where applicable to the case of AI. 4. Actively seek to expand the range of stakeholders and domain experts involved in discussions of these challenges. Happy to be part of the AI Safety Initiative of the Cloud Security Alliance
Electrical Engineering and Systems Science
arxiv.org
To view or add a comment, sign in