The algorithm hides confidential information among a huge number of decoys.
Hackers are constantly improving in ways to bypass cyber protection in order to steal confidential documents. In this regard, a number of researchers from Dartmouth College have proposed using artificial intelligence algorithms to create deceptive files that would hide real data among a large number of forgeries.
The Word Embedding–based Fake Online Repository Generation Engine (WE-FORGE) algorithm generates fake patents, allegedly under development. According to the algorithm developer VS Subrahmanian, someday AI will be able to “create many fake versions of every document that needs to be protected.”
For example, if attackers were looking for a formula for a new medicine, they would have to analyze a large number of counterfeits for real files. This could mean checking each formula in detail and possibly investing in a few deliberately wrong calculations.
Fake documents created by WE-FORGE can also act as hidden traps. For example, a certain file can warn the security system when accessing it. For this strategy, companies usually used human-created forgeries, but now AI can make a huge number of plausible documents.
The system generates convincing traps by searching the document for keywords. For each term it finds, the AI calculates a list of related concepts and replaces the original term with one randomly selected. As a result, dozens of documents can be created that do not contain confidential information, but at the same time look believable.
Catch up on more stories here
Follow us on Facebook here