Kissimmee, Fla. – As U.S. intelligence agencies increase their reliance on commercial geospatial products and services, cybersecurity becomes a growing concern.
“We create so many innovative solutions here in this country,” Stacey Dixon, U.S. deputy director of national intelligence, said May 6 at the 2024 GEOINT Symposium here. “Yet we lose a lot of that innovation to adversaries because we aren’t properly protecting it. Invest in your cybersecurity. Understand that people are going to try to steal your secrets.”
Potential adversaries will seek access to commercial products directly as well as through intermediaries.
Figure out how to prevent that, “because we don’t want our American-made products or our western-made products falling in the hands of adversaries who will use them against us, use them against our troops, use them against our interests,” Dixon said. “Please take the time to invest and know who is buying those capabilities.”
Artificial Intelligence
Information security is also essential as government agencies turn to artificial intelligence for help combining myriad datasets.
“If we have algorithms that can get into the datasets that are currently held in different organizations and agencies and run all of that data at the same time, it’s going to be more effective and provide different kinds of insights than if you do these things in parallel and then combine them at the end,” Dixon said.
Creating those algorithms will be a challenge. In addition, U.S. intelligence agencies seeking to combine datasets with those of other agencies may need to adopt new policies. Legislation may also be required.
“We were created as a community with particular agencies with particular disciplines,” Dixon said.
Laws describe what type of data each agency acquires.
“We’re going to have to figure out how to work through that to still protect the American public while enabling more people to have access to the data,” Dixon said.
Verify and Validate
Before intelligence agencies can trust algorithms and automation, they will need to understand how machines are coming up with the information, Dixon said. For large-language models, for example, “let’s be able to trace it back. That is possible.”
Given increasing misinformation and disinformation campaigns, training analysts to verify and validate information sources will also be essential, Dixon said.
Analysts who develop forecasting tools, for example, will need to prove the value of the tools.
The best way to test a forecasting tool “is for you to tell me something that’s going to happen in the future,” Dixon said. “Then let’s to see whether your prediction was correct.”
When models prove themselves, analysts will adopt them because the models will reduce the amount of time analysts must devote to mundane, repetitive tasks, Dixon said.
“It allows them to use their creative brains to augment what the machines are telling them,” she added.
Related
Source link
lol
