A TechCrunch article published on April 24 announces a new Google product, le Cloud Security AI Workbench, a cybersecurity suite powered by a specialized security AI language model called Sec-PaLM. Reaction from Benoît Grunemwald, cybersecurity expert at Eset France.
From Google's PaLM model, Sec-PaLM is “fine-tuned for security use cases”, explains Google, and incorporates security information such as software vulnerability research, malware, threat indicators and behavioral profiles of threat actors. Cloud Security AI Workbench encompasses a range of new AI-powered tools, like Mandiant's Threat Intelligence AI, which will rely on Sec-PaLM to find, identify and act on security threats. VirusTotal, other property of Google, will use Sec-PaLM to help subscribers analyze and explain the behavior of malicious scripts. otherwise, Sec-PaLM will help Chronicle customers, Google's Cloud cybersecurity service, to search for security events and interact “conservatively” with the results. Users of Google’s Security Command Center AI will get “human-readable” explanations of attack exposure thanks to Sec-PaLM, including affected assets, recommended mitigation measures and risk summaries for safety conclusions, compliance and confidentiality.
Essential human supervision
For Benoît Grunemwald, “There is no doubt that AI/ML plays a vital role in managing everyday cybersecurity challenges. We have been using it for over twenty years. With today's computing power and advances in understanding and integrating AI and ML into virtually every technology field, it is completely natural that these models allow treatments in record time. It should be remembered, however, that all training data and resulting algorithms are reactive and build on what has been seen before.. “Predictive” AI is certainly a promising technology, but it can potentially lead to a large number of false positives, which would lead to the opposite effect of the initial objective. We must continue to rely on skilled human cybersecurity oversight to verify AI results and validate its actions. As in all cutting-edge fields, AI provides essential help, but its implementation and management requires multiple and new skills, both in AI and in the discipline to which it is applied. Our sector is in tension, human resources are scarce, the challenge lies in attractiveness and training”, adds Benoît Grunemwald.
