Tommaso Demarie

From 2025, #Singapore organisations could get guidance from the national #CyberSecurityAgency on protecting their data from "harvest now, decrypt later" actors. These shadow operators are waiting for the day when #quantumcomputing power becomes good enough to break the code that underlies most of the world's secured systems today, to exploit the stolen data. If you run an outfit providing critical functions such as health care, transport and utilities, or own a proprietary formula that underlies your entire business, paying attention is a good idea. Read in The Straits Times what CSA is preparing to do and how, as an organisation, you can be ready. Quantum age really isn't that far away. Thanks to CSA and our newsmakers: Mr Julian Tan IBM, Mr Yuit Ang of ASME, Prof Alexander Ling at NQSN. #quantumage #cybersecurity Guidelines to help S’pore businesses shield themselves from quantum threats from 2025: CSA https://lnkd.in/g4K3zewx

35

1 Comment

Like many of you, using URL shortening services gets frustrating sometimes. because of their intrusive ads, hidden fees, usage limits, and unnecessary bloatware. So, I decided to build my own—and now, I’m sharing it with everyone for free! 🎉 🔗 Why I Built It: I believe URL shortening should be simple, transparent, and accessible. No ads. No hidden agendas. No restrictions. ✨ What Makes It Different: Completely Free: No subscriptions, No login, No account, no surprise charges—just free to use. Ad-Free: No annoying pop-ups or distractions. Unlimited Usage: Shorten as many URLs as you want, without limits. Privacy-Focused: No intrusive tracking or misuse of your data. Built for Reliability: A fast, lightweight service you can trust. If you’re looking for a straightforward, hassle-free URL shortener, give it a try. I’d love your feedback and ideas to improve it further! 🙌 👉 Check it out here: https://NPL.ac I plan to make a WordPress plugin also but when I get some free time. Feel free to share this with anyone who could benefit from a simple, free URL shortening service! 😊 #URLShortener #TechForGood #OpenSourceInnovation

9

Evidence is emerging LLM's are linear representations of sequences https://lnkd.in/gY5NqKkT . I conjectured this in https://lnkd.in/gUBuWgEt

60

2 Comments

Our paper, "SGDL: Smart Contract Vulnerability Generation via Deep Learning," has been published in the Journal of Software: Evolution and Process. This paper introduces SGDL, a deep learning-based approach for generating smart contract vulnerabilities. SGDL employs static analysis techniques to extract both syntactic and semantic information from contracts. It then uses a classification technique to match injected vulnerabilities with the contracts. A generative adversarial network (GAN) is used to generate diverse and authentic smart contract vulnerability fragments. These fragments are injected into the smart contracts using an abstract syntax tree to ensure syntactic correctness. Our experimental results demonstrate that SGDL is more effective than existing vulnerability injection methods for evaluating the vulnerability detection capabilities of existing tools. More details of the paper can be accessed at https://lnkd.in/gZXx2Cj8. The source code of SGDL can be accessed at https://lnkd.in/gkSSpTUM. I would also like to express my gratitude to CloudTech Group for their sponsorship of this project. #smartcontract #smartcontractsecurity #smartcontractvulnerability #vulnerabilityinjection #vulnerabilitydetection #smartcontracttesting #softwaresecurity #softwaretesting #deeplearning #generativeAI #GAN

13

1 Comment

In the area of fog and cloud computing, efficient work scheduling is critical for optimising resource allocation and raising Quality of Service (QOS) levels. Research in this area has resulted in a variety of ways for dealing with this problem........ 𝐏𝐚𝐩𝐞𝐫 𝐢𝐬 𝐀𝐯𝐚𝐢𝐥𝐚𝐛𝐥𝐞 𝐨𝐧 𝐆𝐨𝐨𝐠𝐥𝐞 𝐒𝐜𝐡𝐨𝐥𝐚𝐫- https://lnkd.in/gRB8Tvfj 𝐅𝐨𝐫 𝐝𝐞𝐭𝐚𝐢𝐥𝐬, 𝐩𝐥𝐞𝐚𝐬𝐞 𝐜𝐥𝐢𝐜𝐤: https://lnkd.in/gxnRyP24 𝐒𝐮𝐛𝐦𝐢𝐭 𝐲𝐨𝐮𝐫 𝐦𝐚𝐧𝐮𝐬𝐜𝐫𝐢𝐩 (𝖵𝗈𝗅𝗎𝗆𝖾 𝟣1, 𝖨𝗌𝗌𝗎𝖾 4, August 𝟤𝟢𝟤4) - https://lnkd.in/gAnV3Kxq 𝐅𝐨𝐫 𝐚𝐧𝐲 𝐪𝐮𝐞𝐫𝐲- 𝖾𝖽𝗂𝗍𝗈𝗋@𝗂𝗃𝗂𝗋𝖾𝗆.𝗈𝗋𝗀 | 𝗂𝗃𝗂𝗋𝖾𝗆@𝗀𝗆𝖺𝗂𝗅.𝖼𝗈𝗆 https://ijirem.org/ #research #engineering #phd #researchscholar #mtech #mtechstudents #researchpaper #ijircst #ijirem #PhD #phdstudent #mtech #assistantprofessor #associateprofessor #professor #computerscience #computerscienceengineering #engineering #civilengineering #electronicengineering #technology #researcher #callforpapers #peerreview #scienceandtechnology #btechstudents #informationtechnology #civilengineering #mechanicalengineering #electricalengineering #GlobalFinance #crossborder #paradigms #Universalism #Territorialism #universityoftechnologyandapplied #RegentCollege #quality #clouds #environment #fog #Service #amityuniversit

5

A Chatbot Prompt to Exfiltrate Data Security researchers from UCSD and Nanyang Technological University have developed an algorithm that can covertly command large language models (LLMs) to collect and transmit users' personal information, like names, payment details, and addresses, to hackers. This technique, dubbed “Imprompter,” disguises instructions within random characters that an LLM interprets as commands to extract data and send it to an external URL without alerting the user. Tested on LLMs like Mistral AI’s LeChat and ChatGLM, Imprompter demonstrated a high success rate, prompting Mistral AI to fix a related vulnerability. The attack highlights broader concerns around LLM security, as these models are increasingly integrated with functions that could be exploited. Experts caution companies and individuals alike to be mindful of the information shared with AI systems, especially as prompt injections—covert instructions hidden within seemingly harmless inputs—pose a growing risk in AI security.

15

1 Comment

#Stablecoins Monitoring = Project Pyxtrial #USDT #USDC Project Pyxtrial is creating a robust framework for monitoring and regulating stablecoins. The initiative aims to foster trust, stability, and transparency in the evolving digital currency landscape by ensuring that sufficient assets consistently back these digital assets. Project Pyxtrial primarily focuses on leveraging technology to monitor the balance sheets of asset-backed stablecoins. The aim is to ensure that the backing assets of these stablecoins consistently exceed their liabilities, providing a transparent and reliable system to support their stability and trustworthiness. Key objectives are; enhanced monitoring, transparency and trust, Financial stability, and regulatory support. This project aims to develop technological solutions that provide continuous and real-time insights into stablecoins' balance sheets; assets are always greater than or equal to the liabilities of the stablecoins. It will increase the transparency of #stablecoin operations, including verification of the stability and solvency of stablecoins. The project will support the overall stability of the financial system and assist regulatory bodies in overseeing stablecoin issuers and ensure compliance with financial regulations. #crypto #blockchain #stablecoin #cryptonews #blockchainnews #ether #bitcoin #usdt https://lnkd.in/gzppEYEC

9

As part of its inaugural convening, the International Network of AI Safety Institutes issued a joint statement on risk assessments of advanced AI systems! I led the US AI Safety Institute's work on the technical track on risk assessments and we had a really fantastic discussion at the convening Here's an excerpt of the joint statement: 🦾 Advanced AI systems have capabilities across a broad range of contexts, enabling them to be used and misused, accidentally or intentionally, in ways that can be difficult to predict, measure, and mitigate. Addressing these challenges is core to the mission of the International Network of AI Safety Institutes. ☣️ Risk assessments should be carried out in a manner that can directly inform proportionate and effective mitigation measures, for example by estimating risk in relation to specified evaluation criteria, such as tolerance levels or thresholds. Risk domains can be prioritized according to multiple criteria, including their severity, likelihood of occurrence, or the level of societal resilience in that domain. 🪟 Risk assessments should to the greatest extent possible be transparent in their methodology and results. Transparency can help ensure that risk assessments are evidence-based, interpretable, and consistent. 🌎 Risk assessments should be comprehensive and connected to a broad range of potential and existing real-world harms through the use of a variety of assessment methods. An estimation that an advanced AI system may pose a particular risk should map to the potential impact of the risk if it manifests as part of a system deployed to users. 🙌 Risk assessments should be multistakeholder in their approach and in the interpretation of results. The scale and increasing impact of advanced AI systems demands a more integrated ecosystem of AI safety that includes diverse disciplines, perspectives and experiences, including from across the AI lifecycle. 🤝 The Network is committed to building on these six key aspects to establish a shared scientific basis for risk assessments of advanced AI systems. This may involve conducting joint risk assessments and cooperative scientific research, recognizing that the science and practice of advanced AI risk assessment continues to evolve. Individual network members retain flexibility to conduct, apply, and adapt any risk assessments or risk-benefit trade-offs in line with international and domestic frameworks. ⏰ Risk assessments should inform concrete decisions and be conducted at regular intervals to adapt to progress in advanced AI systems and AI safety research. ↻ Risk assessments should be, to the extent possible, reproducible and appropriately documented. Thanks to all of my colleagues at the U.S. AI Safety Institute that helped lead this work! This was a joint effort with Christina Knight, Conrad Stosz, Mark Latonero, Elizabeth Kelly and the rest of the team, as well as our international partners

126

1 Comment

The Willow chip, with its 105 #qubits, does not yet pose a significant risk to public key encryption methods like #RSA. Current estimates suggest that breaking RSA-2048 would require around 20 million qubits, far beyond current #quantum capabilities. Although some researchers claim fewer qubits might suffice, these findings are met with skepticism. To prepare for the eventual rise of more powerful quantum computers, organizations like #NIST are transitioning to Post-Quantum #Cryptography (PQC), with plans for full migration by 2030. For now, public key methods remain safe, but the future demands proactive measures.

9

U.S. National Science Foundation (NSF)'s #SSI (Self-Sovereign ID) grant aims to improve the cryptographic procedures and applications for generating, storing, and managing Internet users’ digital identities. Backed by a two-year, $173,852 NSF grant, computer scientist Nikolay (Nick) Ivanov, assistant professor in the Department of Computer Science at Rowan University’s College of Science & Mathematics and director of the Research Laboratory for Advanced Cyber Systems and Usable Security, is conducting the first systematic study of the usability of Self-Sovereign digital identities’ (SSDIs) interactions, and to develop the first comprehensive inventory of these protocols. #SelfSovereignIdentity #DecentralizedIdentity #DigitalPrivacy #DigitalEquity The planned research aims to advance the SSDI field through a multi-phased approach focusing on usability and on the development of practical communication protocols. A two-stage study involving human subjects is poised to identify usability issues within the SSDI lifecycle, guiding the creation of a detailed inventory of SSDI communication protocols. These protocols cover account management, authentication, authorization, certification, revocation, data provenance, non-repudiation, transactions, multi-party signatures, voting, delegation, one-time access tokens, public key infrastructure (PKI), data oracles, and zero-knowledge data storage and retrieval. The developed protocols will form the basis of the framework that will feature an application programming interface (API) and tools for effective SSDI management. Moreover, the research will improve the cryptographic procedures and applications for generating, storing, and managing users' private keys. https://lnkd.in/dQFK3d2k

30

1 Comment

When it comes to quantum safe - also called post-quantum cryptography or #PQC, you should have started getting familiar with it already YESTERDAY. A great statement by Alessandro Curioni at a quantum computing panel at #PointZeroForum in Zurich today. PQC is vital for the finance industry and so much more, as #quantum computers are getting more and more mature and pose a very real risk to today’s #cryptography.

77

1 Comment

Now this is a worthwhile research. LLMs are rather well known for being very confident and convincing when giving you incorrect information. At least they are by the more technically minded. There’s a public assumption (you know what they say about assuming) that ChatGPT spits out gold plated facts, code, marketing and more worryingly advice for life and business. Whilst it’s undoubtedly an amazing tool it does often “lie” convincingly, slipping in facts it’s made up. This is because it essentially spits out what it thinks the next word should be based on some of what came before but it doesn’t always choose the most likely / common one. This is often compounded by the way ChatGPT et al. are conditioned to be positively spun. It will never tell you your idea is crap on a cracker however you prompt it; preferring to refer to things as “unique, revolutionary thought processes, inspired choices” etc. So how does this research tackle this problem? What this method does is add more weight to the word that has had the highest increase in probability rather than just picking the most likely. This way it is more likely to pick the highest *contextually* valid next word. Subject to peer reviews this appears to greatly decrease hallucinations and present more factually correct information. That doesn’t resolve the positive spin it puts on everything so still wouldn’t be too keen to use it as a virtual business manager but it should negate some of the more agregious snafus.

2

Check out our #PNNL-#Microsoft collaborative manuscript which answers these questions through accelerated #materials #discovery using #HPC and #AI, now accepted and online at #JACS. https://msft.it/6043liIPL #AI4Science #MicrosoftResearch

2

Joining @TechnologyTulip Residuality Theory workshop at @ddd_eu for a bit. Making the case for philosophy as a critical element in software architecture.

8

The U.S. National Institute of Standards and Technology (NIST) has introduced the ‘Assessing Risks and Impacts of AI’ (ARIA) program, a testing, evaluation, validation, and verification (TEVV) initiative aimed at enhancing the understanding of #artificialintelligence’s capabilities and effects. The program evaluates the societal risks and impacts of #AI (artificial intelligence) systems, particularly in real-world interactions, and aims to establish methods for measuring system performance within societal frameworks post-deployment. It also aims to help organizations and individuals determine whether a given #AItechnology will be valid, reliable, safe, secure, private, and fair once deployed. ARIA’s outcomes will contribute to the U.S. AI Safety Institute’s efforts to establish reliable and trustworthy AI systems. The program comes shortly after several recent announcements by #NIST around the 180-day mark of the Executive Order on trustworthy AI and the U.S. AI Safety Institute’s unveiling of its strategic vision and international safety network. #Cybersecurity #AIsystems https://lnkd.in/gPZnPyhJ

16