The evolution of Sn grain microstructure during thermomechanical cycling occurs due to grain boundary movement, as some single crystal solder joints gradually transform to multicrystals. This process has been documented statistically using polarized light and EBSP mapping by counting the numbers of joints with different orientations and microstructures in packages with different aging and thermomechanical cycling histories. The process by which evolution occurs is was examined assessed by comparing diffraction patterns and peak shifts obtained from in-situ measurements from the full volume of a solder ball at different stages of thermal cycling history using synchrotron radiation. Diffraction patterns were used to identify crystal orientations present in the joint, and peak shape and peak shifts were used to assess changes in internal stress in several joints.
Segmentation has been a foundational concept in network security for decades, evolving from coarse techniques such as virtual local area networks to more granular approaches like micro-segmentation. Yet the absence of clear definitions and a standardized taxonomy has hindered consistent understanding, effective implementation, and rigorous research. This paper introduces practical definitions for key segmentation concepts and a comprehensive taxonomy for classifying methods across technologies, infrastructures, and enforcement strategies. We apply this language to describe major product categories, demonstrate its utility through a real-world enterprise case study, and use it to analyze prior research to identify under-explored areas for future work. Overall, this paper systematizes fragmented segmentation practices by providing clear definitions and a unified taxonomy that support consistent analysis of segmentation design across diverse networked environments.
Investors and analysts spend an average of three minutes and forty-four seconds reviewing startup pitch decks, a process that requires significant manual effort to extract key performance metrics. This challenge is particularly acute for sustainability startups, where both financial and impact indicators aligned with the United Nations Sustainable Development Goals (SDGs) must be evaluated. Addressing this gap, the present study introduces a Generative AI (GenAI)-driven framework that automates the retrieval of financial and sustainability indicators from unstructured pitch deck data. The system integrates OpenAI’s embedding models, FAISS (Facebook AI Similarity Search), and LlamaIndex to enable semantic retrieval and contextual extraction of financial metrics such as revenue and profit/loss, alongside sustainability-related impact narratives. A dataset of 94 Indian sustainability startup pitch decks was analysed to test the framework’s accuracy and efficiency across diverse sectors including clean energy, agritech, and circular economy ventures. Results demonstrate that the model maintains analytical precision and interpretability significantly reducing manual screening time. Furthermore, the framework successfully maps extracted impact indicators to relevant SDGs, allowing for a holistic assessment of financial performance and sustainability outcomes. The findings underscore the transformative role of GenAI and natural language processing in automating due diligence and investment analysis, offering a scalable, transparent, and data-driven approach to evaluating startups. By bridging financial analytics with impact measurement, this study contributes to advancing the use of AI in sustainable finance and provides a foundation for more consistent, efficient, and equitable startup evaluations that align investment decisions with broader societal goals.
This review aims to analyze the changing causes of data breaches overtwo decades by synthesising evidence from various data breachinvestigation reports and regulatory filings. The methodology involvesexamining trends in threat actors, actions, and motives identified inreports such as the Verizon Data Breach Investigations Report (DBIR)series from 2008 to 2024, California Attorney General's reports, and thePrivacy Rights Clearinghouse. (1,2,3) The findings reveal an evolutionthrough distinct phases: an initial period (roughly 2008-2010)dominated by external breaches leveraging hacking and malware, asubsequent era (2011-2019) marked by the rise of sophisticatedcybercrime, including increased phishing and the emergence of definedincident patterns, and a more recent epoch (2020-2024) characterisedby a significant surge in ransomware attacks, exploitation ofsystemic vulnerabilities, and the convergence of financially motivatedand nation-state actors. Throughout these periods, human factors anderrors have consistently contributed to successful breaches. In conclusion, the landscape of databreaches have shifted from simpler external attacks to more complex anddisruptive campaigns, where human vulnerabilities remain a keyenabler, and the emerging landscape includes AI-driven threatsthat are being explored by both attackers and defenders, necessitatingcontinuous adaptation of defence strategies to address both traditionalweaknesses and novel AI-related risks.
At present, the lack of data exchange and sharing between different hospitals leads to the formation of data silos. Additionally, regional healthcare data contain a significant amount of sensitive patient information, and the public access, sharing, and circulation of this data can result in malicious tampering, theft, misuse, and loss of ownership, ultimately compromising patient privacy. Due to the massive volume and unstructured nature of medical data, it becomes more challenging to defend against and trace targeted malicious attacks, such as data theft, tampering, and ransomware attacks. To address these issues, this paper proposes a blockchain-based electronic health record (EHR) sharing scheme with multi-level access control to ensure secure and privacy-preserving data sharing. The proposed scheme incorporates Post-Quantum Attribute-Based Encryption (PQ-ABE) to handle attribute management across decentralized authorities while resisting collusion attacks among users with different permissions. Additionally, Verifiable Secret Sharing is utilized to protect attribute privacy by splitting attributes into attribute names and attribute values, ensuring confidentiality. This paper creates an access policy update algorithm that adds policy blocks across a blockchain to generate a trackable record of policy changes using blockchain transparency, immutability, and audibility features. This approach enables distributed and trustworthy access control management under concealed policy conditions, ensuring data privacy protection and traceability of user actions. Security analysis and experimental results demonstrate that the proposed scheme effectively protects attribute privacy while reducing computational overhead.