Hybridan Research: Sector Review: Artificial intelligence in UK cybersecurity: the human uplift
Hybridan Research: Sector Review: Artificial intelligence in UK cybersecurity: the human uplift
Artificial intelligence (AI) is a double-edged sword in cybersecurity. Whilst new AI models, architectures, and innovations are emerging to protect the security posture of organisations, attackers are also benefiting from deepfakes, sophisticated phishing, and automation of malicious codes. To ensure the impact of AI on cybersecurity to be a net-positive, we need to pit good AI against bad AI.
Point solutions enhanced with machine learning: Global cybersecurity has been built with point solutions, i.e., a single software application that serves a defined purpose (e.g., endpoints, Distributed Denial-of-Service (DDoS)). As the digital infrastructure expands in scale and scope, the deployment of best-in-class point solutions has often resulted in silos and overlaps in cybersecurity coverage. With machine learning (ML) in operation to identify increasingly frequent threats, high-level clarity of the cyber posture is essential to prioritise actions. Security and cybersecurity posture refers to an organisation's overall state of cybersecurity readiness.
Self-learning AI: We believe self-learning AI is unique globally as its AI learns about data from where data sits in organisations and understands the context and the patterns of data on the fly in order to correlate, analyse, prompt alerts and take actions when necessary. This is in stark contrast with the mainstream ML approach of training big data from different sources, which is then cleaned and summarised.
Generative AI (GenAI): Code automation with human alignment promises a productivity gain for both cyber offenders and defenders. NCC Group estimates that a productivity gain of 10% to 30% is achievable throughout a typical software development project. The fine tuning of foundation models for speed is another use case. (Crossword) plans to launch its pen testing service based on LLMs by combining services such as technical due diligence.
Data governance and provenance: GenAI is not built with security-in-design and in fact, AI systems tend to leak their own training data. The governance of data and the prevention of data poisoning attacks (malicious contamination of data to compromise the performance of AI) are critical.
Frontier AI: As companies such as OpenAI Inc and Anthropic PBC continue to push the frontier of general-purpose AI models and pursue AGI (artificial general intelligence, AI capable of solving complex tasks with human cognitive capabilities), practically all industries need to re-configure and reinvent themselves to work with AI and learn to harness the emerging AI capabilities. The cybersecurity industry is no exception.
Read or download the full report here....
Disclaimer & Declaration of Interest
The information, investment views and recommendations in this article are provided for general information purposes only. Nothing in this article should be construed as a solicitation to buy or sell any financial product relating to any companies under discussion or to engage in or refrain from doing so or engaging in any other transaction. Any opinions or comments are made to the best of the knowledge and belief of the writer but no responsibility is accepted for actions based on such opinions or comments. Vox Markets may receive payment from companies mentioned for enhanced profiling or publication presence. The writer may or may not hold investments in the companies under discussion.