Cybersecurity in the age of generative AI: A practical guide for IT experts

Generative AI or GenAI provides efficiency, but also poses additional risks to cyber security. As traditional measures are gradually losing relevance, a new approach is required.

Generative AI offers many benefits in terms of productivity, efficiency and information availability. This technology has the potential to simplify access to knowledge and make it easier for users at all levels - from students and researchers to government officials - to obtain relevant information, understand correct procedures and make informed decisions.

With the spread and integration of these systems in all areas of life security issues are becoming increasingly complex. For this reason, business leaders and cybersecurity professionals should adapt their skills and strategies to better protect data in the digital age.

How generative AI is changing cybersecurity 

The GenAI revolution can lead to undesirable outcomes, such as the proliferation of malicious and unscrupulous generative AI models. Since the launch of ChatGPT in 2022, it has been gradually integrated into most fields - from education to development and entertainment. Since then, however, many copies of the program have surfaced and some have been published on the dark web to usher in a new era of malicious AI-driven attacks:

• Given the success of GenAI, we can expect attacks to accelerate
• Exploiting the potential of this technology goes far beyond human detection and reaction capabilities
• According to the Secureworks Counter Threat Unit, cybercriminals use AI to deploy ransomware within a day of the first intrusion into a company
• This time is steadily decreasing; in 2022, for example, it was 4.5+ days

Most organizations are not yet able to keep up with these threats. Many do not even invest in the most cost-effective and efficient cybersecurity solutions, such as network security monitoring tools or encryption tools, such as VPN-based network protection. Considering that global spending on security solutions and services surpassed the $200 billion mark in 2023, more than 6 million records were exposed globally due to data breaches in the first quarter of 2023 alone.

Traditional methods in cyber security are limited 

On the other hand, the integration of AI models into a company's work requires an expansion of protective measures against attacks. Most cyber security experts today rely on fixed concepts such as patching, firewalls and monitoring. There is no doubt that these methods have their merits and are still absolutely necessary.

However, with the emergence and development of generative AI and its implementation in the work of companies, they are no longer sufficient. GenAI models are dynamic and adaptable, which makes it difficult to protect them using conventional methods.

Just as a person can be manipulated into revealing confidential information, AI models can be tricked and used for malicious purposes. One common type of attack, for example, is a prompt attack.

The static approach of traditional cyber security measures is not suitable for countering these dynamic threats. Accordingly, top managers should consider the possibility of training IT staff in new security options.

Generative AI could serve as a firewall 

However, not everything is so negative. With certain knowledge, language itself can be used as an additional layer of protection in the AI model. Given the vulnerabilities of these systems, which can trigger attacks or be susceptible to linguistic manipulation, this is an excellent solution:

• One of the first defensive measures is the careful creation of meta or system prompts
• These are instructions that control the behavior of the generative AI
• By properly crafting these prompts, cybersecurity professionals can limit the scope of AI responses, reducing the risk of exposing sensitive information

TipA well-constructed metaprompt can be designed to politely reject all requests that are intended to extract confidential data or elicit inappropriate responses.

Setting up prompt and response filtering for generative AI

Another measure is the implementation of a separate AI model that checks both input queries and the generated output for controversial material. This is a kind of filtering of what goes into the system and also a careful examination of the responses it generates.

For example, if generative AI only responds to public queries, it could have a discrete AI associated with it that blocks (and ideally records) any generated responses that could be considered controversial or harmful. 

By using voice as a firewall, organizations can add an additional layer of security that is well equipped to meet the challenges of new AI-based technologies. This approach ensures that both input and output are controlled and filtered by equally powerful tools, providing more comprehensive protection against traditional and emerging cyber threats.

Image from Pixabay

Further opportunities for generative AI in cybersecurity

Generative AI in Security Operations Centers (SOCs) and Security Event and Incident Management (SEIM) is essential to improve cybersecurity and reduce threats. In a SOC, AI models can identify patterns that indicate cyber threats that traditional detection systems may miss:

• Ransomware
• Malware
• unusual network traffic

Generative AI enables more sophisticated data analysis and anomaly detection in SIEM systems. By learning from historical security data, AI tools can establish a baseline of normal network behavior and detect deviations that may indicate security incidents.

Advantages of AI models for cybersecurity

Generative AI improves the ability to effectively detect and mitigate cyber threats. Using deep learning models, AI can simulate complex attack scenarios that are critical for testing and improving security systems. This modeling capability will help develop robust defenses against known and emerging threats.

As cyber threats become more complex, the proactive and adaptive nature of generative AI becomes increasingly important for maintaining the integrity and resilience of cybersecurity infrastructures.

The right use of AI can shape the future of cyber security

Cybersecurity is one of the most important areas of application for generative AI. In this direction, the power of GenAI works in two ways: it is a powerful tool for those who commit cybercrime and an equally powerful tool for cybersecurity professionals responsible for preventing and/or mitigating the risk of cybercrime.

With the continuous expansion of the functions of generative AI models, the introduction of new protective measures and the use of GenAI technologies to implement them is unavoidable. Companies should consider investing in the development or acquisition of new AI-based security tools in good time, as well as targeted training of IT staff on the possibilities of this technology.