Linkedin, September 1, 2022
We see how machine learning is used in our life: it is implemented almost in all industries. Solutions based on machine learning shape our lifestyle. However, along with their expansion, the number and severity of security problems are increasing.
In order to study machine learning algorithms from an attacker’s perspective, there is a special area of machine learning research called Adversarial Machine Learning (AML). Scientists in this domain study threat models, most modern attacks and countermeasures in an ongoing arms race.
Lukas Bieringer, Kathrin Gross, and other researchers conducted a qualitative study. The purpose was to find answers to the following questions:
Read the article with the results of scientist’s work at the link.
EAIDB, August 27, 2022
Daily media strips are full of information and news about incidents related to privacy violations, algorithmic biases, and AI omissions. Society has already moved on from a state of total oblivion. It has come to the recognition that AI technologies and the huge amounts of data on which they are trained already pose a real threat to privacy, accountability and transparency, and fair society.
In this regard, the demand for ethical AI services is also growing. Ethical AI is gaining popularity among many parties in the innovation environment from startup leaders developing AI-based solutions to corporate clients implementing such solutions, as well as society at large.
The Ethical AI Database (EAIDB) project, which aims to take a fundamental step from understanding existing and potential problems to learning possible solutions, released a new report last week. EAIDB is developed in collaboration with the Ethical AI Governance Group and is an in-depth market research for a growing ethical AI startup dedicated to adopting the responsible AI development, deployment, implementation and management.
EAIDB defines an “ethical AI company” as a company that either provides the methods and tools to make existing AI systems ethical, or one that creates products that remove elements of bias, unfairness, or “unethicalness” in society. The number of such companies has risen significantly over the past few years and the new ones continue to appear.
IEEE Spectrum, August 29, 2022
Facial recognition technology is spreading but it remains vulnerable to attacks. That is why a group of researchers is calling on hackers worldwide to enter a competition aimed at identifying flaws in facial recognition systems and raising awareness of existing and potential risks.
Adversa AI knows firsthand how vulnerable machine learning models are as the company regularly conducts Red Team studies to test AI systems for weaknesses. Adversa AI has developed a challenge for this competition, in which hackers can change photos of celebrities so that the facial recognition system identifies them incorrectly.
According to Adversa AI CTO Eugene Neelou, there are more and more tools for carrying out attacks, there are already real cases. He said – “It’s very easy and fast for attackers with enough motivation. Our engagements show that some of the best facial-recognition vendors demonstrate little to no security against adversarial input modifications”.
The organizers plan that this competition will highlight the current topical issues related to face recognition systems. The winner will also be required to publish the methods used to help the industry close potential flaws. The essence of the task is that the participant with a set of 10 photos of famous celebrities and online access to a face recognition model trained to recognize them, tries to slightly change the images so that the model fails to identify them. The competition is held from August 12 till September 23.
CBInsights, August 31, 2022
New technologies have entered our lives and are steadily changing them and affecting all industries. But the extent to which they are able to solve the most difficult problems of mankind depends on the direction and realization of their potential.
The technologies that are most relevant for the next 10-20 years were presented in the Game Changers 2022 report. Among these technologies, AI auditing occupies a leading position, namely platforms that test AI models to protect against risks such as cyberattacks and hidden biases.
Stay up to date with what is happening! Get a first look at news, noteworthy research and worst attacks on AI delivered right in your inbox.
NIST’s image standards could become win or lose factor for gov biometrics contractors BiometricUpdate, August 22, 2022 The face image quality standard, which could be in place by early 2024, ...
