When spam meets AI
- By Anoushka Deshmukh
- Jul 31, 2019
Even as spam continues to flood email inboxes, it is morphing as it spreads across websites and messaging applications. It can take the form of fake product reviews designed to damage a competitor or a complex network of spam bots used to polarize an entire nation over social media. And while it is now taking advantage of artificial intelligence to create more deceptive communications, spam can seriously compromise other algorithms, according to Emilio Ferrara, writing in Communications of the ACM.
AI relies heavily on the training data it is provided. If any spam, meaning unwanted information, were injected into training data, then the AI system could be misled and form falsely supported conclusions that impact its behaviors. These spamming tactics could cause the misreading of medical scans or the misdirection of autonomous vehicles.
Spam is "likely to plague our society and our systems for the foreseeable future," Ferrara said, because it "feeds itself of (economic, political, ideological, among others) incentives and of new technologies, both of which there is no shortage of."
To combat new forms of spam, he offers three recommendations:
Design secure technology. New technologies should be designed with the possibilities of exploitation in mind so they can be equipped to deal with threats.
Play the long game. Fighting spam is an arms race so researchers must anticipate the need for countermeasures when spammers abuse their designed technologies.
Implement blockchain technology. Blockchain’s authentication mechanism can prevent spam bots from passing as verified users and keep them out of networks.
Read the full paper here.
Anoushka Deshmukh is an intern with Public Sector 360, writing for GCN, FCW and Defense Systems.