Intel announced it secured a four-year contract to work on a Defense Advanced Research Projects Agency (DARPA) cybersecurity project last week. The chipmaker, in tandem with the Georgia Institute of Technology, will develop tools to protect machine learning (ML) programs against deception attacks.
The Santa Clara, California-based manufacturer said it would initially work on improving DARPA ML models’ semantic, spatial, and temporal coherence.
Making Machine Learning Programs More Secure
Increasingly, technology companies are using ML applications to expand their computer systems capabilities beyond their programming. Whereas traditional platforms can only address problems according to their design, ML programs can integrate new information to refine their approach. For instance, Tesla’s Autopilot system uses ML to recognize objects like speed limit signs and driving lanes.
However, despite their advanced functionality, ML programs are not foolproof and can be subverted by malicious operators. Last April, a group of Chinese data scientists guided a Tesla Model S into making a near-fatal lane change by tricking its object recognition system with a series of stickers.
To prevent ML models from being similarly tricked, DARPA started an initiative called Guaranteeing AI Robustness against Deception (GARD). The program’s mandate is to train advanced computing systems to handle attacks outside their design parameters. Fourteen months after launching GARD, the government has brought in Intel to enhance its ML models.
Why Intel Won the GARD Contract
Although Intel is best known as a manufacturer of laptop and server processors, the firm has taken a significant interest in the AI sector in recent years.
Last December, the corporation acquired Habana Labs, a startup that made chipsets capable of accelerating AI functionality.
In January, the manufacturer showcased the latest innovations of its Mobileye subsidiary at CES 2020. The segment demonstrated it had trained its self-driving AI to navigate through dense city traffic seamlessly.
Moreover, Intel revealed it developed an intelligent computing system with Cornell University that can replicate a mammal’s sense of smell.
Because of its experience in the field and cutting-edge hardware, Intel has the resources DARPA needs to ensure GARD’s success. Also, the manufacturer can utilize its Defense Department contract funding to further its AI research and development.