cybersecurity

The big picture: AI model operators don’t have a good way of reigning in these malicious use cases, Kayne McGladrey, a senior member of the Institute of Electrical and Electronics Engineers (IEEE), told Axios. Allowing LLMs to digest and train on CVE data can help defenders synthesize the wave of threat alerts coming their way each day. Operators have only two real choices in this type of situation: allow the models to train on security vulnerability data or completely block them from accessing vulnerability lists, he added. “It’s going to be a feature of the landscape because it is a dual-use technology at the end of the day,” McGladrey said.

Kayne McGladrey, IEEE Senior Member, noted that the use of generative AI models in business hinges on their ability to provide accurate information. He cited as examples studies of AI models’ abilities to extract information from documents used for financial sector regulation that are frequently relied on to make investment decisions. “Right now, the best AI models get 80 percent of the questions right,” McGladrey said. “They hallucinate the other 20 percent of the time. That’s not a good sign if you think you are making investment decisions based on artificial intelligence telling you this is a great strategy four out of five times.”

Algorithmic bias is one of the primary risks associated with emerging physical surveillance technologies. While the risks of facial recognition software are well known and documented, efforts are being taken to adapt computer vision to new and novel use cases. For example, one of the more deeply flawed failures was an attempt to detect aggressive behaviour or body language, which was unfeasible as there was not enough training data available. Other physical security systems will face a similar challenge of not discriminating against individuals based on protected factors due to a lack of training data, or more likely, a lack of gender or racially unbiased training data. Companies considering purchasing advanced or emerging physical security systems should enquire about the training data used in the development of those systems to not be subject to civil penalties resulting from discrimination caused by using said systems.

CISOs now face substantial personal risks, as seen in cases like Uber and SolarWinds where the SEC has taken legal action against the security chiefs. The primary risk is both personal and professional liability for the CISO, according to Kayne McGladrey, field CISO at Hyperproof. The problem, however, is that boards unaware of the business risks from poor cybersecurity are unlikely to include the CISO in the Directors & Officers insurance policy. “This exposes CISOs to substantial risk,” McGladrey told Cybersecurity Dive.

Prepare to laugh until your stomach hurts with our most hilarious episode yet, featuring the one and only theater kid turned cybersecurity guru, Kayne McGladrey, Field CISO at Hyperproof. Join us for a rollercoaster of emotions as we dive into the absurdity of security info in 10K filings, engage in heated debates over the polarizing cinnamon sticky bun ale, and champion the cause for more singing and dancing in cybersecurity. Think of it as the “Cybersecurity’s Got Talent” episode you never knew you needed! Kayne’s journey is packed with invaluable insights and captivating stories that are as unique as they are engaging.

Despite this guidance mandating only four disclosures (identifying and managing risks, disclosing material breaches, board oversight, and management’s role), over 40% of the 2,100+ 10-K filings I’ve reviewed between January 1 and March 11, 2024 disclosed eleven distinct topics.

Companies are disclosing more information than required in their 10-K filings for various reasons. One is that they lack a broad consensus how much detail to disclose in Section 1C. The recent civil litigation of SEC vs. Tim Brown and SolarWinds (case 1:23-cv-09518 in the Southern District of New York) significantly influences the disclosure requirements.

Tom provides an update on the status of the Hyperproof FedRAMP project. Along the way, Kayne uncovers some of the challenges associated with the project and suggests solutions for others going through the same process. And straight out of left field, Kayne actually likes a beer more than Tom. Come find out what caused this seismic disturbance in the force.

In recent years, we’ve seen a significant shift in the threats targeting businesses. “Everybody focused on the human harms, people couldn’t check into their hotel rooms; people couldn’t use an ATM… the nature of the technical exploits is not what we focus on in terms of harm… that’s not what we focus on in terms of harm,” states Kayne McGladrey, a field CISO at Hyperproof and senior IEEE member. This reiterates the transition from mere inconvenience to significant operational disruptions and economic consequences that cyber threats now pose.

During this Hyperproof live stream series, leaders in information security shed light on crucial topics that shape the modern cybersecurity landscape. This month’s episode features Jeff Warren, Owner & Principal Consultant at South Lake Cyber Risk, LLC, and our host, Kayne McGladrey, Field CISO at Hyperproof. Guided by Kayne and audience questions, Jeff will share insights into his current work and past experiences in the field. Register now for your chance to learn from one of today’s top infosec pros.

Kayne McGladrey, Field CISO at Hyperproof and IEEE Senior Member, noted that the use of generative AI models in business hinges on their ability to provide accurate information. He cited as examples studies of AI models’ abilities to extract information from documents used for financial sector regulation that are frequently relied on to make investment decisions.

“Right now, the best AI models get 80 percent of the questions right,” McGladrey said. “They hallucinate the other 20 percent of the time. That’s not a good sign if you think you are making investment decisions based on artificial intelligence telling you this is a great strategy four out of five times.”

Join Tom Coughlin, J.L. Doty, Gloria Washington, Kathleen Kramer and me as we explore the jobs we’ll see created over the next 20 years, industry’s role driving innovation and the skills our future workforce will need.

“These threats are not merely theoretical, although, at the moment, they are still relatively limited in their application,” McGladrey said. “It is reasonable to expect that threat actors will continue to find innovative new uses of generative AI, extending beyond business email compromise, deepfakes, and the generation of attack code.”