Connect with us

Technology

Google Cloud’s security chief warns: Cyber defenses must evolve to counter AI abuses

Published

on

Google Cloud brings tech behind Search and YouTube to enterprise gen AI apps

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More


While many existing risks and controls can apply to generative AI, the groundbreaking technology has many nuances that require new tactics, as well. 

Models are susceptible to hallucinations, or the production of inaccurate content. Other risks include the leaking of sensitive data via a model’s output, tainting of models that can allow for prompt manipulation and biases as a consequence of poor training data selection or insufficiently well-controlled fine-tuning and training. 

Ultimately, conventional cyber detection and response needs to be expanded to monitor for AI abuses — and AI should conversely be used for defensive advantage, said Phil Venables, CISO of Google Cloud.

Advertisement

“The secure, safe and trusted use of AI encompasses a set of techniques that many teams have not historically brought together,” Venables noted in a virtual session at the recent Cloud Security Alliance Global AI Symposium.

Lessons learned at Google Cloud

Venables argued for the importance of delivering controls and common frameworks so that every AI instance or deployment does not start all over again from scratch. 

“Remember that the problem is an end-to-end business process or mission objective, not just a technical problem in the environment,” he said. 

Nearly everyone by now is familiar with many of the risks associated with the potential abuse of training data and fine-tuned data. “Mitigating the risks of data poisoning is vital, as is ensuring the appropriateness of the data for other risks,” said Venables. 

Advertisement

Importantly, enterprises should ensure that data used for training and tuning is sanitized and protected and that the lineage or provenance of that data is maintained with “strong integrity.”

“Now, obviously, you can’t just wish this were true,” Venables acknowledged. “You have to actually do the work to curate and track the use of data.”

This requires implementing specific controls and tools with security built in that act together to deliver model training, fine-tuning and testing. This is particularly important to assure that models are not tampered with, either in the software, the weights or any of their other parameters, Venables noted. 

“If we don’t take care of this, we expose ourselves to multiple different flavors of backdoor risks that can compromise the security and safety of the deployed business or mission process,” he said. 

Advertisement

Filtering to fight against prompt injection

Another big issue is model abuse from outsiders. Models may be tainted through training data or other parameters that get them to behave against broader controls, said Venables. This could include adversarial tactics such as prompt manipulation and subversion. 

Venables pointed out that there are plenty of examples of people manipulating prompts both directly and indirectly to cause unintended outcomes in the face of “naively defended, or flat-out unprotected models.” 

This could be text embedded in images or other inputs in single or multimodal models, with problematic prompts “perturbing the output.”

“Much of the headline-grabbing attention is triggering on unsafe content generation, some of this can be quite amusing,” said Venables.

Advertisement

It’s important to ensure that inputs are filtered for a range of trust, safety and security goals, he said. This should include “pervasive logging” and observability, as well as strong access control controls that are maintained on models, code, data and test data, as well. 

“The test data can influence model behavior in interesting and potentially risky ways,” said Venables. 

Controlling the output, as well

Users getting models to misbehave is indicative of the need to manage not just the input, but the output, as well, Venables pointed out. Enterprises can create filters and outbound controls — or “circuit breakers” —around how a model can manipulate data, or actuate physical processes. 

“It’s not just adversarial-driven behavior, but also accidental model behavior,” said Venables. 

Advertisement

Organizations should monitor for and address software vulnerabilities in the supporting infrastructure itself, Venables advised. End-to-end platforms can control the data and the software lifecycle and help manage the operational risk of AI integration into business and mission-critical processes and applications. 

“Ultimately here it’s about mitigating the operational risks of the actions of the model’s output, in essence, to control the agent behavior, to provide defensive depth of unintended actions,” said Venables. 

He recommended sandboxing and enforcing the least privilege for all AI applications. Models should be governed and protected and tightly shielded through independent monitoring API filters or constructs to validate and regulate behavior. Applications should also be run in lockdown loads and enterprises need to focus on observability and logging actions. 

In the end, “it’s all about sanitizing, protecting, governing your training, tuning and test data. It’s about enforcing strong access controls on the models, the data, the software and the deployed infrastructure. It’s about filtering inputs and outputs to and from those models, then finally making sure you’re sandboxing more use and applications in some risk and control framework that provides defense in depth.”

Advertisement

Source link
Advertisement
Continue Reading
Advertisement
Click to comment

You must be logged in to post a comment Login

Leave a Reply

Technology

Meta’s new OS update for Quest includes a redesign and train mode

Published

on

Meta’s new OS update for Quest includes a redesign and train mode

Meta is introducing some big changes with its Quest v71 update, including a redesign of Meta Horizon OS, a calendar app, and the ability to use Travel Mode on a train. The update will start rolling out gradually next week.

Let’s start with the redesign. Meta says that it’s “tweaking the look and feel of Horizon OS” and that new headsets will use an “improved” light theme by default. (Dark mode fans, fear not: the dark theme is still present, and if you’re already using it, Meta won’t switch you over automatically.) Meta has “also made assorted changes to the way panels look and behave, where the control bar resides, the colors of various UI elements, and so on,” according to a blog post. And the Settings menu has “a new look and layout” and improved search.

An image of the updated settings menu.
Image: Meta

The calendar app looks like, well, a calendar app, and based on a brief video, it appears you’ll be able to look through different days while you’re in VR. Meta says the app can integrate with Google and Outlook calendars and will also include Meta Horizon Worlds events you subscribe to and meetings you’re invited to that take place in Meta Horizon Workrooms.

Advertisement

With the update to Travel Mode, you’ll be able to use the feature when on trains. Meta first introduced the feature in May, but at the time, it was only available to use on airplanes. Note that when in Travel Mode, you can’t use the Quest Pro or Touch Plus controllers, so some games might be difficult to play. Instead, Travel Mode is probably better suited for watching a movie or TV show while in transit. (Meta also explicitly says that you should not use travel mode in a car.)

Source link

Continue Reading

Technology

Between Perplexity’s new macOS app and ChatGPT’s search launch, conversational search just got a lot more fun

Published

on

Perplexity AI

The AI-powered search game is heating up, with OpenAI adding search capabilities to ChatGPT just yesterday, and Perplexity launching a macOS desktop app a week ago. Perplexity could be considered a leader in AI-powered search right now, and now Mac users can quickly use Perplexity’s advanced search engine right from their desktop without having to access it through a web browser.

If you’re not familiar with Perplexity, it’s a conversational search engine that launched the same year as ChatGPT. It allows you to ask questions and make queries using natural language (like how we communicate with each other in real life). Similar to ChatGPT, there are two tiers for users – a free ‘Standard’ tier which allows you to make unlimited ‘Quick’ searches, ideal for those seeking quick, basic answers produced by the Standard Perplexity AI Model (plus five free Pro tier searches a day), and the ‘Pro’ tier which includes Quick searches as well as 600 Pro searches a day.

Source link

Continue Reading

Science & Environment

Dominion is discussing small nuclear reactors with other tech companies

Published

on

Dominion is discussing small nuclear reactors with other tech companies


The Dominion coal burning power plant is seen in Saint Paul, Virginia on Tuesday, February 7, 2023. Community leaders in Southwestern Virginia are giving serious consideration to the idea of utilizing formerly mined coal sites to house small modular nuclear reactors.

Mike Belleme | The Washington Post | Getty Images

Dominion Energy is talking with other tech companies about developing small modular nuclear reactors, after the Virginia utility entered into an agreement with Amazon last month to look at advancing the next generation technology .

“It’s very encouraging to see large power users, including technology companies, express a willingness to invest, partner and collaborate to bring this exciting base load carbon free technology into fruition,” Dominion CEO Robert Blue told investors on the company’s third-quarter earnings call Friday.

Advertisement

Dominion and Amazon have signed a memorandum of understanding to explore developing a small modular reactor near the utility’s North Anna nuclear station in Louisa County, Virginia. The small reactor would bring 300 megawatts of power to Virginia.

Virginia is one of the most nuclear friendly states in the nation with strong bipartisan support for next-generation nuclear initiatives, Blue said.

“It’s not surprising that our large customers would be interested as they think about us as a good operator of nuclear, to work together on maybe advancing those kinds of technologies,” the CEO told investors.

“So we’ve been talking with Amazon obviously and others,” the CEO said.

Advertisement

Tech companies are investing in nuclear power as they hunt for carbon-free, reliable electricity to support the growing energy needs of artificial intelligence data centers. Dominion serves the largest data center market in the world, northern Virginia.

Earlier this year, Amazon bought a data center campus from Talen Energy that will be powered by the Susquehanna nuclear plant in Pennsylvania. Microsoft has signed an agreement to purchase power from Three Mile Island as Constellation Energy aims to restart the plant in 2028. And Alphabet‘s Google agreed last month to purchase power from the startup Kairos Power, a developer of small modular reactors.

Small modular reactors promise to reduce capital costs and speed the deployment of nuclear plants. They have a smaller footprint than large reactors, making them easier to site in principle, and promise a simpler manufacturing process.

But the technology has struggled to reach the commercial stage. There is no operating small modular reactor in the U.S. right now.

Advertisement



Source link

Continue Reading

Technology

Play this underrated Marvel game for free with Prime Gaming

Published

on

Play this underrated Marvel game for free with Prime Gaming
Gamora, Groot, Starlord, Rocket Raccoon, and Drax standing ready to fight. Groot is holding a blue llama.
Eidos-Montreal

Amazon Prime Gaming has 24 games up for grabs for members during November, and many are worth your time. But you’ll want to check out the first game on the list, which is now available for free.

Marvel’s Guardians of the Galaxy, developed by Eidos-Montreal, was vastly underrated when it came out in 2021. While it was critically acclaimed, with one of the best superhero narratives we’ve seen outside of the PlayStation Spider-Man games and an excellent 1980s-inspired soundtrack, then parent company Square Enix said it “undershot” expectations. Whether it was due to poor marketing, comparisons to the poorly received live-service game Marvel’s Avengers (also published by Square Enix) from the previous year, or something else entirely, it flew under the radar.

It did manage to eventually surpass 8 million players, according to Crystal Dynamics, thanks to it entering PlayStation Plus and Xbox Game Pass catalogs. Now, you have the chance to get it with another deal through Prime Gaming. At the very least, you should check it out because it has Cosmo, who is the fan-favorite talking dog, and also a llama for some reason.

There are other great games you can grab as well, including Dishonored: Definitive Edition on November 7,  Tomb Raider: Anniversary on November 14, ridiculous couch co-op game Overcooked: Gourmet Edition on November 21, and the space MMO Elite Dangerous on November 27.

You can check out the full list of games below. Be sure to double-check the platform, as some are for GOG, while others are for the Epic Games Store or the Amazon Games App. And as a reminder, you’re automatically a Prime Gaming member if you subscribe to Amazon Prime.

Advertisement

November 1

  • Marvel’s Guardians of the Galaxy (Epic Games Store)
  • Mafia: Definitive Edition (GOG)

November 7

  • Dishonored: Definitive Edition (GOG)
  • Duck Paradox (GOG)
  • Close To The Sun (GOG)
  • Disney Pixar Cars (Amazon Games App)
  • Bang Bang Racing (Amazon Games App)
  • Snakebird Complete (Epic Games Store)

November 14

  • Ms. Holmes: The Case of the Dancing Men (Amazon Games App)
  • Chasm: The Rift (GOG)
  • House of Golf 2 (Epic Games Store)
  • Tomb Raider: Anniversary (GOG)
  • Blade of Darkness (GOG)

November 21

  • Max: The Curse of Brotherhood (Amazon Games App)
  • Overcooked: Gourmet Edition (GOG)
  • Gloomy Tales: One-Way Ticket (Legacy game code)
  • Super Meat Boy (Epic Games Store)
  • Moonscars (GOG)
  • Riot: Civil Unrest (GOG)

November 27

  • Elite Dangerous (Epic Games Store)
  • Sir Whoopa**: Immortal Death (GOG)
  • Jurassic World Evolution (Epic Games Store)
  • Mystery Case Files: The Dalimar Legacy (Amazon Games App)
  • Shogun Showdown (GOG)






Source link

Continue Reading

Technology

TIDAL could lay off a quarter of its current staff

Published

on

TIDAL could lay off a quarter of its current staff

TIDAL, the high-definition music streaming platform, is not having the best of times. The company has already made layoffs, and its management has been mulling a potential restructuring to get back on track to profitability. Now, a report claims that an upcoming round of layoffs will affect around a quarter of the current TIDAL staff.

Competition in the music streaming service segment has become quite fierce. Other platforms have been getting more and more features and a growing catalog of songs. TIDAL’s main selling point over the others has always been Hi-Fi audio. While rival services were betting on formats with quality compression, TIDAL offered premium plans with lossless music.

TIDAL could get rid of a quarter of its staff in new round of layoffs

However, the arrival of high-quality audio plans at rival services probably led a percentage of TIDAL users to consider switching. While the latter’s Hi-Fi quality typically offers a higher bit rate than its rivals, only a specific niche of users truly value this feature. It even requires using earbuds or headphones of a certain range to enjoy it.

In December 2023, the company shed around 10% of its workforce. Now, according to Fortune, an upcoming round of layoffs will impact around a quarter of TIDAL’s workforce. The source says they gained access to a memo from Jack Dorsey, the CEO of Block Block Head (Tidal’s parent company). “We’re reducing the size of our design team and foundational roles supporting TIDAL, and we will consider reducing engineering over the next few weeks as we have more clarity around leadership going forward,” the memo said, according to Fortune.

Advertisement

In this round of layoffs, around 100 employees could leave the company. In July of this year, Dorsey was reportedly considering restructuring TIDAL. The staff reduction could align with this plan. TIDAL is not the only tech company to have made massive layoffs recently. Others, like DropBox and even Meta, have had to do the same.

Recent moves to make TIDAL more attractive

Earlier this year, in a move to become more competitive in the market, TIDAL bundled its Hi-Fi plans. Priced at $10.99, the new premium subscription undercuts the previous, more expensive one. The company has also made moves to solidify its position as a streaming service focused on Hi-Fi audio. The list includes setting FLAC as the default format and adding Dolby Atmos support.

Source link

Continue Reading

Technology

Polestar delivers the first US-made Polestar 3 EVs

Published

on

Polestar delivers the first US-made Polestar 3 EVs

The first American-made Polestar EVs are now on the road. The Swedish automaker said on Friday it delivered the first Polestar 3 SUVs to US and Canadian customers. The Inflation Reduction Act, signed by President Biden in 2022, incentivizes automakers to manufacture EVs in the US, offering tax credits for companies and buyers.

Polestar began stateside Polestar 3 production in August at its Ridgeville, South Carolina plant. Those vehicles will serve North American and European markets, complementing the company’s more established production in Chengdu, China.

The Polestar 3 electric SUV sitting by a lake.

Tim Stevens for Engadget

The automaker says American and Canadian Polestar 3 deliveries will pick up steam in the coming weeks. You can take one for a spin if you live near one of the automaker’s Polestar Spaces, essentially brand-building fancy showrooms in or near major US cities.

The Polestar 3 is billed as the company’s coming-out party, expanding the niche reach of the first two models to a more mainstream audience. Engadget’s Tim Stevens tested the EV last month and found that, apart from a few early software glitches, the $73,400 and up SUV is “great.”

Advertisement

Source link

Continue Reading

Trending

Copyright © 2024 WordupNews.com