31 décembre 2018 | International, C4ISR

US Spies Want to Know How to Spot Compromised AI

BY DAVE GERSHGORN

What if you were training an AI, and an adversary slipped a few altered images into its study set?

The US government's research arm for intelligence organizations, IARPA, is looking for ideas on how to detect “Trojan” attacks on artificial intelligence, according to government procurement documents.

Here's the problem the agency wants to solve: At a simple level, modern image-recognition AI learns from analyzing many images of an object. If you want to train an algorithm to detect pictures of a road signs, you have to supply it with pictures of different signs from all different angles. The algorithm learns the relationships between the pixels of the images, and how the structures and patterns of stop signs differ from those of speed-limit signs.

But suppose that, during the AI-training phase, an adversary slipped a few extra images (Trojan horses) into your speed-limit-sign detector, ones showing stop signs with sticky notes on them. Now, if the adversary wants to trick your AI in the real world into thinking a stop sign is a speed-limit sign, it just has to put a sticky note on it. Imagine this in the world of autonomous cars; it could be a nightmare scenario.

The kinds of tools that IARPA (Intelligence Advanced Research Projects Activity) wants would be able to detect issues or anomalies after the algorithm has been trained to recognize different objects in images.

This isn't the only kind of attack on AI that's possible. Security researchers have also warned about inherent flaws in the way artificial intelligence perceives the world, making it possible to alter physical objects like stop signs to make AI algorithms miscategorize them without ever messing with how it was trained, called “adversarial examples.”

While neither Trojan attacks nor the adversarial examples are known to have been used by malicious parties in the real world, researchers have said they're increasingly possible. IARPA is looking at a short timeline as well, expecting the program to conclude after a maximum of two years.

https://www.defenseone.com/technology/2018/12/us-spies-want-know-how-spot-compromised-ai/153826

Sur le même sujet

  • CACI Awarded $197 Million Task Order to Support Special Electronic Mission Aircraft Flight Operations for U.S. Army

    25 septembre 2019 | International, Aérospatial

    CACI Awarded $197 Million Task Order to Support Special Electronic Mission Aircraft Flight Operations for U.S. Army

    ARLINGTON, Va.--(BUSINESS WIRE)-- CACI International Inc (NYSE:CACI) announced today it has been awarded a more than four-year task order, with a ceiling value of $197 million, to support the U.S. Army Fixed Wing Project Office's special electronic mission aircraft flight operations worldwide. Under the task order, CACI will use its proven technical and training solutions for operational mission support to provide 24/7 surveillance and reconnaissance assistance to U.S. and NATO forces deployed abroad. Awarded under the RS3 contract vehicle, the work represents new work for CACI. John Mengucci, CACI President and Chief Executive Officer, said, “CACI will leverage its successful surveillance and reconnaissance expertise to provide the U.S. Army with comprehensive flight operations support. We are committed to delivering advanced technology and highly qualified personnel to meet our customers' evolving requirements.” CACI Executive Chairman and Chairman of the Board Dr. J.P. (Jack) London, said, “CACI is prepared to advance and support the critical surveillance and reconnaissance capabilities the U.S. Army needs to accomplish its mission and safeguard our nation.” CACI's 22,000 talented employees are vigilant in providing the unique expertise and distinctive technology that address our customers' greatest enterprise and mission challenges. Our culture of good character, innovation, and excellence drives our success and earns us recognition as a Fortune World's Most Admired Company. As a member of the Fortune 1000 Largest Companies, the Russell 1000 Index, and the S&P MidCap 400 Index, we consistently deliver strong shareholder value. Visit us at www.caci.com. There are statements made herein which do not address historical facts, and therefore could be interpreted to be forward-looking statements as that term is defined in the Private Securities Litigation Reform Act of 1995. Such statements are subject to factors that could cause actual results to differ materially from anticipated results. The factors that could cause actual results to differ materially from those anticipated include, but are not limited to, the risk factors set forth in CACI's Annual Report on Form 10-K for the fiscal year ended June 30, 2019, and other such filings that CACI makes with the Securities and Exchange Commission from time to time. Any forward-looking statements should not be unduly relied upon and only speak as of the date hereof. CACI-Contract Award View source version on businesswire.com: https://www.businesswire.com/news/home/20190925005100/en/ Corporate Communications and Media: Jody Brown, Executive Vice President, Public Relations (703) 841-7801, jbrown@caci.com Investor Relations: Daniel Leckburg, Senior Vice President, Investor Relations (703) 841-7666, dleckburg@caci.com Source: CACI International Inc https://www.epicos.com/article/486889/caci-awarded-197-million-task-order-support-special-electronic-mission-aircraft

  • Northrop Grumman delivers mini-laser system to US Armed Forces

    28 août 2023 | International, C4ISR

    Northrop Grumman delivers mini-laser system to US Armed Forces

    Known as the ‘Phantom’, this mini laser system will be integrated into a wide range of tactical solutions across the American armed forces.

  • Navy seeks to offer virtual training to more of the fleet

    27 août 2023 | International, Naval, C4ISR

    Navy seeks to offer virtual training to more of the fleet

    The Navy's live, virtual and constructive (LVC) tech investments support everything from massive Large Scale Exercises to routine pre-deployment training.

Toutes les nouvelles