Show Notes
Artificial intelligence is often described as a "black box". We can see what we put in, and what comes out. But not how the model comes to its results. And, unlike conventional software, large language models are non-deterministic. The same inputs can produce different results. This makes it hard to secure AI systems, and to assure their users that they are secure. There is already growing evidence that malicious actors are using AI to find vulnerabilities, carry out reconnaissance, and fine-tune their attacks. But the risks posed by AI systems themselves could be even greater. Our guest this week has set out to secure AI, by developing red team testing methods that take into account both the nature of AI, and the unique risks it poses. Peter Garraghan is professor at Lancaster University, and founder and CEO at Mindgard. Interview by Stephen Pritchard
More from Security Insights
Businesses are capturing, and storing, ever larger volumes of data. These data sources are vital for decision making, and of course for AI. But they can pose security risks. Data can be corrupt, contain malware, or even just be wrong. And as enterprises connect more systems to shared data -- in data warehouses or datalakes -- any threat can quickly spread right across the business. But, as our guest for this episide argues, often security isn’t even in the room when it comes to discussions about data management. Even so, security expertise is vital, if we are to ensure that new data projects do not create new vulnerabilities, epecially in a data driven enterprise. Our guest is is Danielle Kinsella, senior direct at Gigamon. Interview by Stephen Pritchard.
In the face of growing cyber threats, the obvious step is to ask for more budget. But does spending more on security tools always improve an organisation’s defences? Are there more effective ways for cybersecurity teams to marshal their resources? This is where security frameworks come in. Standards such as TOGAF, Open FAIR, zero trust architecture and an upcoming framework on security roles from the Open Group are all vital tools to help CISOs define what they need to defend, and how. Our guest is John Lindford, Forum Director of The Open Group’s Security Portfolio.
A recent survey suggests that, in the UK at least, boards overestimate their organisations' ability to withstand a cyber attack. And they overstate their ability to recover from an incident, and restore even to minimum viable operations. What does that mean for security, and does it increase risk? And will pressure, from regulators, law makers, and even shareholders and investors, force boards to take cyber resilience more seriously, and view it at least on par with business continuity? Our guests are Paul Cragg, CTO at NormCyber, and Richard Beeston, Chief Operating Officer at Digital Space.
Recent incidents show the risks AI can pose to cybersecurity. At the same time, the security industry is putting a lot of store on AI as a way to bolster defences. So is AI a threat, or an answer to our security challenges? As the CEO of a business supplying security services to global organisations, our guest is well placed to discuss how CISOs view the risks, and benefits of artificial intelligence. And, as Raluca Saceanu, CEO of Smarttech247 puts it, there are those who are "all in" on AI, and those who view it with distrust. Interview by Stephen Pritchard.
Physical and cyber security are separate disciplines. But are they converging, and what can they learn from each other? Do changes in the threat landscape, from geopolitics to AI, mean physical and cyber teams need to work more closely together? And can a more joined-up approach to security improve operational resilience? Our guest is Cohesity's James Blake.