Vijay’s Quora session was focused on voice security, machine learning, social engineering, as well as the Atlanta tech scene. Here are the top five questions and answers from Friday’s session:
1. How will applications of machine learning change in the next 10 years?
With machine learning, we are heading towards a future where technology becomes more human. We are already seeing this trend with machine learning getting close to human parity in tasks like image recognition and classification, speech recognition, self driving cars…(cont.)
2. How is social engineering being used as a hacking tactic?
Social engineering is about building trust through careful exchange of information, such that the person on the other end of the informational transaction is eventually convinced to perform some task.Hackers know that it is much easier to hack a human than a machine…(cont.)
3. How can consumers ensure that their voice-controlled products are secure?
I would recommend that the consumers look at the following to ensure that their voice-controlled products are secure: Is it equipped with trustworthy voice biometrics? While some products have voice biometric technology already integrated, the technology by itself is often vulnerable to malicious voice attacks such as pitch morphing or replay attacks. If your device includes voice biometric technology, make sure that it also includes voice spoofing countermeasures…(cont.)
4. What are the most recent developments in the field of voice security as of 2017?
There have been many interesting developments in the field of voice security this year. One is the availability of secured voice-activated IoT devices. At the beginning of this year, there were several news stories about accidents involving Amazon Echo and Google Home that raised security concerns. While the current solutions are still not optimal, good progress has been made towards making voice interactions more secure…(cont.)
5. What are the shortcomings of machine learning?
Machine Learning has several shortcomings. First ML is only as good as the quality of data used in the models. The old adage, “Garbage in, garbage out” holds true here. In addition, with ever increasing amounts of data and model complexity, it is easier than ever to reach false conclusions or “see what you want to see” when developing ML models…(cont.)
Visit Vijay’s Quora session here to read the full answers and insights on machine learning and the future of voice.
