“All things are numbers,” professed Pythagoras. Today, twenty-five centuries later, algebra and mathematics are everywhere in our lives, whether we see them or not. Artificial intelligence (AI) explosions like the Cambrian have brought numbers even closer to all of us, as advances in technology have made it possible to parallelize vast amounts of operations.
Over time, operations between scalars (numbers) were parallelized into operations between vectors, followed by matrices. Multiplication between matrices is currently trending as the most time- and energy-intensive operation in modern AI computing systems. A technique called “tiled matrix multiplication” (TMM) helps speed up computation by breaking matrix operations into smaller tiles to be computed by the same system in successive time slots. However, modern electronic AI engines using transistors are approaching their inherent limits and can hardly compute at clock frequencies above 2 GHz.
Light’s compelling qualifications for lightning speed and significant energy and footprint savings provide a solution. Recently, a team of photonics researchers from the WinPhos Research group, led by Professor Nikos Pleros of Aristotelian University in Thessaloniki, harnessed the power of light to create a compact silicon photonics computer that can compute his TMM with his record-breaking 50 GHz clock. developed the engine. frequency.
as reported in Advanced photonics, which employs a silicon-germanium electro-absorption modulator and a novel neuromorphic architecture design capable of encoding and computing data. According to corresponding author George Giamougiannis, “This work paves the way for solutions for DL-based applications that require line rate calculations.” This work promises to make a significant contribution to the data center. cyber security.
Data Center Cybersecurity: The Light That Hunts Evil
Arguably, AI Burst has equipped both benign and malicious users with a powerful toolkit to accelerate and automate their activities.with data movement data center (DC) is growing by about 13% year-over-year, making it a prime target for malicious individuals looking to compromise. confidential data,for example, financial data, personal informationand intellectual property of many organizations including: Government agency, the military, hospitals, and financial institutions. DC cybersecurity is therefore essential to prevent intruders from accessing sensitive information.
In fact, threat detection mechanisms are facing a set of new requirements due to the large amount of data flowing through the vast number of servers and switches within a modern DC. Real-time threat detection is essential. Packet inspection must be processed at lightning speed. Additionally, threats should be detected as early as possible in the path of malicious packets. Every DC node should be equipped with a strong cybersecurity toolkit.
Leveraging their ultra-fast processors, researchers at Aristotle University of Thessaloniki, in collaboration with NVIDIA experts in the DC cybersecurity field, have successfully combined silicon photonics and AI to develop the most common types of DCs. We have established a framework to quickly and successfully identify one of our attacks. A line-rate distributed denial of service (DDoS) attack on NVIDIA’s servers. Thanks to this novel calculation scheme, the number of DC attacks could soon increase, at least for the time being.
For more information:
George Giamougiannis et al., Neuromorphic Silicon Photonics with 50 GHz Tiled Matrix Multiplication for Deep Learning Applications, Advanced photonics (2023). DOI: 10.1117/1.AP.5.1.016004
Quote: Performing Matrix Multiplication at the Speed of Light for Enhanced Cybersecurity (Feb 1, 2023) Feb 2, 2023 https://techxplore.com/news/2023-02-matrix-multiplications-cybersecurity Taken from .html
This document is subject to copyright. No part may be reproduced without written permission, except in fair trade for personal research or research purposes. Content is provided for informational purposes only.