Note: This article is written by me with a DOI: 10.13140/RG.2.2.34423.94881

Abstract-We discuss three crucial points that can be understood as AI Security. However, we focus more on the least explored use of the term “AI Security.” Below, we outline three kinds of uses of the phrase “AI Security.” With the growth of AI at an unprecedented speed, we need to understand new issues that have not existed before. For example, we never envisioned our online AI avatars and cyberattacks on these avatars. We never imagined that cyber attackers would use GenAI to compromise our systems. Providing solutions through AI to these problems is a key area of concern in modern cybersecurity. Hence, there is a need for AI-based cybersecurity, both for securing systems that utilize AI and for protecting systems from AI-based cyberattacks.

Introduction

With the growth of technologies come significant problems that can emerge if some screws are left loose. With the modernization and development we achieve come challenges that need to be addressed. One such key concern is cybersecurity and the safety of AI. This can be seen as security by AI, security for AI, and security to the AI. We need to recognize that with this much growth in AI, our systems can become vulnerable.

To explain, if we use a Chinese app for AI, all data of your AI Avatar on this site is in China. Therefore, another Chinese company can hack your data and access your AI Avatar. Your data is now visible to the hackers. They can use it in various ways. The attacker may edit your AI preferences, delete your AI Avatar, or copy your AI Avatar for research or other uses, such as country-based analysis or more. Other points include helping cyber attackers detect hacking with the help of AI. AI has reached such a pinnacle of development that it can be used to avoid hacking. AI can detect cyber crimes and help protect systems from cyber attacks. In the same vein, AI can create intelligent bugs that can attack systems; we need to protect our systems from all these things that are byproducts of modernization and the development of technology.

Three points can be understood from the term “AI Security”. Here are the points:

1. AI security as AI-powered security engine.

2. Secondly, AI security involves AI-powered cyberattacks on vulnerable systems.

3. Thirdly, AI security is needed to protect users’ AI avatars.

These will be explained in the following sections.

I. AI security for AI avatars

To secure user information and data on the servers of AI companies, crucial steps must be taken. This data comes from AI individuals and is stored in massive data centers and corresponds to a person’s AI avatar. It must be protected. If this data is lost, three things can happen: (1) the data privacy of affected individuals would be compromised, posing a threat to their online accounts; (2) attackers can manipulate the data and (3) data can be copied. Cyber attackers can use people’s AI avatars to impersonate them, access various intelligence channels, and disrupt systems. Therefore, the security of AI avatars is of utmost importance. User data must be safeguarded. For instance, if ChatGPT is a competitor of a Chinese company, the loss of AI personal data could affect many people in countless ways that cannot be adequately expressed in mere words. An attacker may edit your AI preferences, delete your AI avatar, or copy your AI avatar for research or other purposes, such as country-based analysis. Thus, the aim should be to protect the AI avatars of users in AI systems.

II. AI security as an AI-powered cyberattack on vulnerable systems

Now, AI can be used to attack other AI or non-AI-based systems. This is different from the above. In the previous case, we say that the AI information of a person needs security. Here, we say that AI itself can be used to attack someone’s system. Given that AI has become intelligent, manipulators can access other systems using AI. AI-based viruses can enter other networks and steal data, manipulate it, or disrupt the systems. This needs to be addressed. We may require newer antivirus solutions that can detect AI-based viruses or malware. These can be embedded in images and videos of GenAI products. One must have updated protections. For that, antivirus companies should be able to detect when an AI-based virus has attacked the system. Thus, the detection of AI-based viruses can be achieved with the help of AI, as explained in the section below.

III. AI security as an AI-powered security engine.

When the security of a system is at risk, AI can help. In the case discussed above, when AI-based attacks on a system can occur, AI can assist as well. As I always mention, there are two things — good AI and bad AI. So here, let’s discuss the good AI fighting non-AI-generated or AI-based viruses: the bad AI. AI can indeed provide robust security to systems, where we can use Machine Learning and Natural Language Processing to detect and stop cyber attacks. Here, AI can be used to analyze the knowledge base to make logical conclusions, indicating that the path of traversal on a router can lead to a corrupt file or virus in some way. So AI can really help in this case. Prior data on security breaches from companies like Facebook should be shared with international cybersecurity teams so that the patterns followed by cyber attackers are well known and understood. Future missions of cyber attackers can be anticipated by GenAI, LLMs, and ML. We can avoid future attacks by understanding; just like the black box of a plane, past cyber attacks can be analyzed and mitigated for a safer future to stop cyberattacks on both AI and non-AI systems alike.

Summary

In summary, AI can help ensure that people and AI Avatars are safe from attacks on the internet. These attacks may be intended to steal information or cause systems to malfunction, depending on the hacker’s aim. We can study previous security breaches to learn how to prevent new ones. Another use of the term AI security is to protect systems with the help of AI. The third use of the phrase AI security refers to safeguarding systems from AI-based attacks, which are attacks where the attacker uses AI to enhance the attack. There is a lot that needs to be done. We will discuss more in upcoming articles.