Latest news with #CamelliaChan
Forbes
2 days ago
- Business
- Forbes
20 Tech Experts On Emerging Hardware Trends Businesses Must Watch
While emerging technologies like artificial intelligence and blockchain have been capturing headlines, businesses must also keep an eye on evolving hardware trends. As companies pursue faster, smarter and more secure operations, hardware configuration is becoming a critical competitive differentiator. In an era shaped by AI adoption, rising end-user expectations and tightening privacy regulations, IT leaders are reevaluating not only what hardware their organizations need, but also where it should reside and how it should be deployed. Below, members of Forbes Technology Council share key hardware strategies designed to deliver the flexibility, security and cost efficiency modern enterprises require. 1. AI-Embedded Hardware Security At The Edge AI-embedded hardware security at the edge is becoming essential. By integrating intelligent processing directly into devices—servers, endpoints and storage—companies can achieve real-time, autonomous security; reduce latency; and protect privacy without cloud dependence. This hardware-native AI trend will be critical for secure, scalable operations in the near term. - Camellia Chan, Flexxon 2. Inference-Optimized Hardware We're seeing a shift toward inference-optimized hardware—systems designed specifically for running, not training, AI models. As model deployment scales, general-purpose GPUs waste energy and rack space. Purpose-built accelerators with high memory bandwidth utilization will be essential for cost-effective, real-time AI. - Thomas Sohmers, Positron AI Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify? 3. Disaggregated Infrastructure Disaggregated infrastructure is rising fast—separating compute, storage and memory lets companies scale AI workloads efficiently. Paired with smart NICs and GPUs, it's the backbone for low-latency, high-throughput architectures in tomorrow's data centers. - Sai Krishna Manohar Cheemakurthi, U.S. Bank 4. Field-Deployed Edge AI Accelerators Edge AI accelerators are gaining traction—particularly in the insurance industry, where devices that can analyze claims locally are deployed in field adjusters' kits. These devices slash cloud costs while preserving privacy. Key benefits include TOPS/watt efficiency, hardware-encrypted data pipelines, and precertification for IEC 62304 medical-grade reliability. The future is distributed intelligence. - Srinath Chandramohan, EY 5. Hybrid Cloud Infrastructure Hybrid cloud—combining locally hosted servers and public cloud providers—is a trending configuration. This strategy can help businesses reduce costs while maintaining the flexibility to scale when needed. - Anto Joseph, Eigen Labs 6. Edge-Enabled Safety Infrastructure As a public safety company, we're seeing increasing demand for edge-enabled safety infrastructure—devices like smart panic buttons, mobile gateways and compact edge processors that can locally process video, audio or wellness data before syncing with cloud-based platforms. This reduces latency, enables real-time decision-making in emergencies, and enhances privacy by limiting data exposure. - Kevin Mullins, SaferMobility 7. AI-Enabled Edge Computing A key trend is integrating edge computing with AI, enabling real-time data processing near the source. This reduces latency and bandwidth, which is crucial for IoT and smart systems. Advances in processors and AI chips facilitate local data analysis, enhancing decision-making and efficiency. Adopting this trend can help companies in data-driven industries cut costs, improve performance and stay competitive. - Gautam Nadkarni, Wipro 8. Localized Foundation Model Deployment Enabling hardware to run localized foundational models is key. Current foundation models and large language models require significant infrastructure and pose security risks due to generalization and centralized data processing. New approaches deploy personalized models on small devices like watches or phones, enabling secure local processing and paving the way for personalized AI assistants. - Abhijeet Mukkawar, Siemens Digital Industries Software 9. On-Site Edge Computing In Factories And Telecom Sites A growing trend is building edge computing setups equipped with GPUs or AI chips close to where data is generated. The advantages include reduced delays, because information is processed locally; bandwidth savings; and scalability. These reliable, flexible and modular hardware configurations allow factories and telecom sites to run AI-powered tasks on site, enabling faster responses, stronger data protection and more efficient workload management. - Maman Ibrahim, EugeneZonda Cyber Consulting Services 10. Data Center Layouts Built For AI Local compute is making a comeback. Everyone chased the cloud—until inference costs punched them in the face. We used to fight over RAM; now it's NVMe lanes, PCIe bandwidth and power delivery. Welcome to the AI hardware wars! But AI-native workloads demand rack design, not just chip choice. If your data center layout hasn't changed since 2015, you're not ready. - Mirror Tang, ZEROBASE 11. Energy-Optimized Hardware We're in the early stages of a shift toward energy-optimized hardware. Organizations are investing in renewable-powered data centers to meet ESG goals and reduce their carbon footprints. - Ohm Kundurthy, Santander Bank 12. Accelerated Compute AI Clusters I'm seeing accelerated compute AI clusters doing double duty for both training and inference. The push toward agentic and multimodal AI requires significant processing power to solve complex problems and advance AI autonomy. - Steven Carlini, Schneider Electric 13. Modular Hardware Setups One significant trend is the shift to modular hardware setups, such as servers and storage, which allow for scaling up or down as needed. This lets companies add power or space without a full rebuild, making it easier to keep up with changing needs and control costs. It's a flexible approach that's quickly becoming standard. - Ganesh Ariyur, Gainwell Technologies 14. Edge Computing With NPUs And Specialized Chips One essential hardware trend is AI-accelerated edge computing. By processing data closer to its source with neural processing units and specialized chips, companies reduce latency, improve privacy and enable real-time decision-making. As AI becomes core to operations, edge intelligence will be critical for speed, scalability and resilience. - Rishit Lakhani, Nile 15. Heterogeneous Hardware Compatibility Adopting heterogeneous hardware compatibility and mixed-hardware serving is essential. This enables the flexible use of diverse hardware types—GPUs, CPUs and ASICs—across generations and vendors, boosting capacity utilization and cutting costs. It supports scalable AI workloads by running models efficiently on mixed hardware fleets, increasing agility and sustainability. - Pooja Jain, Meta (Facebook) 16. Privacy-Driven Edge Computing A growing hardware trend is edge computing—processing data closer to the user instead of relying entirely on the cloud. It's becoming essential for real-time decision-making in privacy-sensitive environments. For example, in AdTech, edge setups enable brands to deliver faster, more compliant, personalized ads without sacrificing speed or data security. - Ivan Guzenko, SmartyAds Inc. 17. Hybrid CPU-GPU Architectures A clear short-term trend is the adoption of hybrid CPU-GPU architectures optimized for AI and data analytics workloads. It's important to understand that AI is no longer optional—it must be integrated into workflows. These architectures improve performance without requiring full infrastructure replacement, helping companies balance cost and efficiency. - David Barberá Costarrosa, Beeping Fulfilment 18. Chip-Level Security Integration A key hardware trend is the integration of security at the silicon level—such as trusted platform modules, secure enclaves and hardware-based authentication. With rising cyberthreats and remote workforces, companies must adopt hardware that enforces zero-trust principles from the chip up to protect sensitive data and systems. - Raj Jhaveri, Greenlane™ Infrastructure 19. On-Device NPUs One essential hardware trend is the adoption of neural processing units on personal devices. Newer PCs and devices come equipped with NPUs to handle AI workloads efficiently—on the device. As AI becomes integral to everyday workflows, devices without these chips risk falling behind in performance and capability. - Tarun Eldho Alias, Neem Inc. 20. Heterogeneous Compute One key trend is the move to heterogeneous compute—combining CPUs, GPUs and AI accelerators—to handle growing machine learning workloads. Traditional CPUs can't keep up with large models. Adopting specialized hardware like H100s, faster interconnects and memory-rich nodes is essential for faster training, cost efficiency and staying competitive in the AI era. - Karan Alang, Versa Networks Inc.
Techday NZ
28-04-2025
- Business
- Techday NZ
X-PHY unveils real-time on-device deepfake detection tool
X-PHY has launched a new on-device deepfake detection tool designed to help users identify manipulated media in real time. The newly released X-PHY Deepfake Detector extends the company's reach beyond traditional hardware-layer data protection, aiming to address the surge in AI-generated fraud and media manipulation worldwide. The solution promises up to 90% accuracy in the identification of AI-generated videos, images, and audio content, and processes data entirely on the user's device without the need for cloud connectivity. The rapid proliferation of deepfake content has been highlighted as a growing global concern. According to X-PHY, deepfake material on social media increased by 550% between 2019 and 2023, with the World Economic Forum identifying it as a significant worldwide risk. X-PHY's latest product seeks to enable users to verify the authenticity of digital media in real time, representing the company's move into AI-driven content integrity solutions and bridging the gap between data protection and digital trust. Camellia Chan, Chief Executive Officer and Co-Founder of X-PHY, stated: "At X-PHY, we are committed to extending our ethos of Security by Design beyond data protection. The X-PHY Deepfake Detector strengthens our vision of a Community Root of Trust, where every layer - from hardware to data to content - serves as a checkpoint for authenticity and security. By combining deepfake detection with our existing hardware-embedded defences, we're ensuring every endpoint not only protects data, but actively discerns and verifies the trustworthiness of the information flowing through it." The X-PHY Deepfake Detector utilises a multi-modal AI approach. Upon activation, it analyses video, image, and audio streams instantly, examining facial micro-expressions, voice fingerprints, and GAN-generated artefacts for indications of manipulation. The detection process operates locally on the user's device, guaranteeing privacy and enabling use even when disconnected from the internet. This is achieved by leveraging advanced temporal and spatial AI analysis with pre-trained neural networks. The tool is designed to identify subtle discrepancies in facial movements, audio waveforms, and image characteristics that are commonly associated with AI-generated content. X-PHY asserts that integrating this technology with its established hardware-based protections results in a cohesive security system, covering both stored data and the authenticity of digital communications. Deployment options for the Deepfake Detector are tailored to support a diverse range of enterprise requirements. It can be installed as a lightweight software agent on Windows-based computers and laptops, or deployed in conjunction with the X-PHY Cybersecure SSD. This approach enables organisations to create a single defence layer that encompasses data protection, ransomware prevention, and deepfake detection. The tool is designed for broad compatibility and integration. Its application-agnostic structure allows for operation alongside widely used platforms, including Teams, Zoom, Webex, Chrome, YouTube, and Meta. Users can activate the detection tool with a single click when entering a meeting, while the software autonomously runs for a preset interval and can be reactivated as required. Built on Zero Trust principles, the X-PHY Deepfake Detector is intended to provide an additional layer of authentication and verification directly at the device, helping to bolster organisations' resilience against AI-driven deception. It also lessens dependence on third-party validation systems, which can sometimes add unnecessary operational complications. The X-PHY Deepfake Detector is now available for purchase through official X-PHY sales channels and authorised global partners.
Yahoo
24-04-2025
- Business
- Yahoo
X-PHY Inc Unveils Real-Time Deepfake Detection Tool Ahead of RSA Conference 2025
Expanding its AI-powered security suite, X-PHY takes aim at the rising threat of AI-generated deception SAN FRANCISCO, April 24, 2025 /PRNewswire/ -- X-PHY Inc, a leading innovator in embedded cybersecurity technology, has announced the launch of its latest solution ahead of RSA Conference (RSAC) 2025 - Deepfake Detector - a real-time deepfake detection tool that empowers users to verify the authenticity of videos, audio, and images directly on their devices, without relying on the cloud. Live demonstrations will be held for the first time during RSAC. The growth of deepfakes has been exponential – deepfake content on social media alone grew 550% between 2019 and 2023, and the World Economic Forum states it is a key global risk. X-PHY's innovation is designed to combat AI-generated deception, enabling users to verify the authenticity of digital media - including videos, images, and audio – with up to 90% accuracy, in real-time. This marks X-PHY's expansion into AI-driven content integrity solutions, bridging data protection with digital trust. "At X-PHY, we are committed to extending our ethos of Security by Design beyond data protection," said Camellia Chan, CEO and Co-Founder of X-PHY Inc. "The X-PHY Deepfake Detector strengthens our vision of a Community Root of Trust, where every layer - from hardware to data to content - serves as a checkpoint for authenticity and security. By combining deepfake detection with our existing hardware-embedded defences, we're ensuring every endpoint not only protects data, but actively discerns and verifies the trustworthiness of the information flowing through it." On-Demand Deepfake Detection Upon activation, the X-PHY Deepfake Detector uses multi-modal AI to analyze video, image, and audio streams in real time. By examining facial micro-expressions, voice fingerprints, and Generative Adversarial Network (GAN)-generated artifacts, it flags signs of manipulation - even across multiple video windows. Detection is performed entirely on-device, preserving privacy and functioning even without an internet connection. This is achieved through the Deepfake Detector's use of advanced temporal and spatial AI analysis, powered by pre-trained neural networks. These models are capable of identifying subtle inconsistencies across facial movements, audio waveforms, and image artifacts - common signs of AI-generated content. When combined with X-PHY's patented hardware-based protections, this forms a seamless security ecosystem, protecting both stored data and the integrity of digital communications. Flexible Deployment, Fuss-Free Integration Designed for seamless adoption, the Deepfake Detector offers flexible deployment options to suit varying enterprise needs. It can be installed as a lightweight software agent on personal computers and laptops running on Windows operating systems or packaged with the X-PHY Cybersecure SSD - creating a unified defense layer that spans data protection, ransomware prevention, and deepfake detection. The solution is application-agnostic, compatible with leading platforms like Teams, Zoom, Webex, Chrome, YouTube, and Meta. Users can activate it with a single click when joining a meeting, where it runs autonomously for a preset duration and can be re-engaged as needed. Built on Zero Trust principles, the solution adds an additional layer of authentication and verification at the device level, helping organizations strengthen their cyber resilience against AI-powered deception and reducing reliance on external validation systems that often introduce unnecessary operational complexity. X-PHY Deepfake Detector is now available for purchase through the official X-PHY website and from authorized global channel partners. For enterprise enquiries or bulk deployments, please contact our sales team at info@ Experience live demonstrations at X-PHY Inc's Booth #5368, located in the North Expo Hall of the Moscone Center, from April 28 to May 1 during RSA Conference 2025. About X-PHY Inc X-PHY Inc is a pioneering cybersecurity company dedicated to hardware-based cybersecurity solutions that protect data at its core. Built on the principle of Security by Design, X-PHY embeds protection directly at the physical layer for proactive, autonomous, and real-time defense against evolving cyber threats. Headquartered in California, USA, X-PHY Inc was established in 2021 and has since developed a growing portfolio of 43 patents, reinforcing its commitment to innovative AI-embedded security at the hardware level. The company's patented solutions safeguard endpoints, servers, and data centers, ensuring zero-trust resilience across industries. X-PHY Inc is part of the Flexxon Group, a leader in hardware engineering and memory solutions, leveraging its legacy of innovation and expertise in secure storage to build cutting-edge cybersecurity technologies for the digital world. For more information, please visit:X-PHY: View original content: SOURCE X-PHY Inc Sign in to access your portfolio
Forbes
09-04-2025
- Forbes
The Cybersecurity Wake-Up Call: Hardware And Trust Are Our Future
Camellia Chan is the CEO and Cofounder of Flexxon, a next-generation hardware cybersecurity solutions provider with a global presence. getty News broke in February 2025 that hackers chained three vulnerabilities in Palo Alto Networks' PAN-OS firewalls, turning a trusted security gatekeeper into an open door. Thousands of unpatched systems fell, exposing sensitive networks to root-level compromise. But before we go placing the blame on the company, it's important to recognize that this was not just a vendor slip-up. It was a flare illuminating a deeper truth: Software-centric security is crumbling under modern threats. To survive, we need to evolve and expand our approach. Strengthening community-rooted trust anchored in hardware—through innovations at the physical layer—must be our next step. The nightmare unfolded with CVE-2025-0108, CVE-2025-0111 and CVE-2024-9474—flaws in PAN-OS that let attackers bypass authentication, read files and escalate to root access via exposed management interfaces. Proof-of-concept code spread quickly, and exploits followed even faster, outpacing organizations' ability to patch. The impact? Firewalls meant to protect became backdoors. This highlights a harsh truth: Relying on software updates alone leaves us forever reacting—often too late. This recent attack isn't an outlier; it's a symptom. In 2020, the attack involving SolarWinds saw hackers lace a software update with malware, hitting 18,000 organizations. This major incident captured headlines and ignited massive downstream repercussions, reinforcing that vendor trust can be a single point of failure. Yet almost five years down the road, here we still are. The 2023 breach involving Barracuda went further. A zero-day attack forced hardware replacements, not just quick software patches, showing that when software fails you can't always patch your way out of the problem. In a 2021 incident, a critical vulnerability named Log4Shell was discovered in a Log4j library. Hackers could remotely run malicious code, and the damage spread like wildfire. These aren't isolated incidents—they reveal a pattern. Patching after the fact isn't enough. We need a foundation that stops breaches before they spread. A community root of trust flips the script. Trust isn't any single vendor's burden. It is a shared fortress built by hardware makers, software developers, researchers and users. Think collective threat intelligence spotting exploits early, or ecosystem-wide standards ensuring devices aren't weak links. Palo Alto's race to patch couldn't match hackers' speed, but a community model could have shrunk that window because shared accountability outpaces solo fixes. Contrast this with today's reality: isolated vendors, running alone, while attackers feast on the gaps. It's time to stop seeing cyber threats as somebody else's problem, it is ours. If software is the lock, hardware is the door. If the door's flimsy, no lock will hold back invaders. The most recent case proves it: Software fell like dominoes because the hardware beneath lacked intrinsic defenses. Hardware-based security is harder to crack remotely, and offering bedrock software cannot match. Imagine if those firewalls had integrity checks built into their silicon. Exploits might have hit a wall before root access was theirs. A more resilient approach must combine hardware-rooted protections that prevent software failures from escalating. Cryptographic roots of trust, embedded in hardware, verify system integrity from the moment a device boots up. Firmware-level security prevents unauthorized modifications, ensuring attackers can't manipulate the system undetected. Beyond these, AI-driven security at the memory level adds another layer of defense, autonomously detecting ransomware and unauthorized access in real time. Unlike software-based monitoring, these AI-enhanced solutions operate within the hardware itself—responding instantly without relying on external updates or human intervention. Establishing trust at the system level begins with ensuring that the foundation—both the hardware and firmware—remains uncompromised. The Trusted Platform Module (TPM) plays a key role here. As a security chip or firmware solution, TPM provides cryptographic verification, ensuring a device's boot process, encryption keys and firmware remain untampered. It acts as a safeguard, preventing compromised software from executing unchecked. However, TPM alone is not a silver bullet. It is part of a broader movement toward hardware-integrated security, working alongside newer advancements such as AI-powered storage security, secure enclaves and tamper-resistant firmware. The goal is to create multilayered trust mechanisms that harden security. This layered approach significantly reduces the attack surface and ensures systems are resilient against both remote and physical exploits. The Palo Alto breach demonstrates why this shift is necessary—software defenses alone are not enough. A strategic mix of TPM, AI-driven security and other hardware protections represents the path forward, ensuring security isn't an afterthought but a built-in standard. So how do we begin fixing this? Community-Driven Trust: Vendors, researchers and users must collaborate. Think of shared threat databases or a 'trust certification' for devices meeting hardware/software benchmarks. Hardware Mandate: Critical devices such as firewalls, routers and servers need security baked in. TPM should be standard, verifying integrity from boot to runtime. AI Hardware Integration: Push for smart AI-embedded security technology at the storage level, paired with TPM for systemwide resilience. A community could standardize this combo, ensuring no layer is left exposed. User Action: Enterprises should restrict management interfaces (as Palo Alto urged), but also demand TPM-enabled devices and verify it pre-deployment. Policy Push: Governments could incentivize hardware security through tax breaks for TPM adoption and penalties for repeat breaches to make resilience a mandate not a cost comparison. Today, we stand on a mountain of major cyber incidents. The recent cybersecurity breaches—they are not just warnings; they're a blueprint for failure or success. Software's fragility demands a community-rooted trust model, reinforced by hardware security. This isn't a solo sprint by any one vendor—it's a collective stand to ensure trust spans the entire ecosystem, not just a single patch. IT pros, policymakers and vendors must act now and build the fortress, making it silicon-strong and community-wide, before the next exploit chain strikes. The future's not secure until we make it so. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
