It is not uncommon for Threat Actors in the Cyber Domain, to use and at the same time abuse the Information Technology/ Communication Technology Infrastructure by means of Artificial Intelligence (AI) and Machine Learning (ML), to carry out Malicious Activities. Technology and Software releases then become the next milestone for the Threat Actors to experiment and find the next Spoke to dent the Infrastructure/Software that is being targeted. Also, in this renewed venture, we find that the Actors gain success, much easier than expected.
What is OpenAI’s “ChatGPT”
The new OpenAI’s new software platform, “ChatGPT” (Generative Pre-trained Transformer) released in the end of November 2022 is a ‘Chatbot’ that uses both AL and ML. This new ChatBot uses an innovative and new interface for its Large Language Model (LLM), which instantly created a flurry of interest in AI and its possible uses. It is built on top of OpenAI’s GPT-3 family of Large language models, and is fine-tuned (an approach to transfer learning) with both supervised and reinforcement learning techniques. Both these approaches that is incorporated in the ChatGPT, uses human trainers to improve the model’s performance. Post its launch, the prototype quickly garnered attention for its detailed responses and articulate answers across many domains of knowledge. As the model launched, is still in the learning phase, it is seen to be still uneven on factual accuracy and this has been identified as a significant drawback, as on date. Further, it is worth a mention that the ChatGPT bot, post its launch is making waves when it comes to all the amazing things it can do—from writing music to coding to generating vulnerability exploits, and others. The Chatbot, within a few days of its launch, surpassed a million users to the extent that its servers could not take on the subscriber load. Also, it is to be understood that, when more and more users subscribe and start submitting their chats, the ‘ChatGPT’ will attain maturity and at the same time, may also bring out the limitations in AI’s thinking, notwithstanding the creators attempting to mend the negative throws, in real time.
Cybercriminals Starting to Use “ChatGPT”
Now, we at “Cyber Secure India” are of the opinion, that this new ChatGPT has added another avenue to the Threat Spectrum in this modern cyber threat landscape to facilitate any Threat Actor to perform malicious activities; as it is apparent that code generation can help less-skilled threat actors, plan and implement Cyber Attacks using the feature. Infact, in the short span of its life cycle through its presence in the Cyber Arena, ChatGPT malicious users, have been able to successfully conducted a full infection flow, from creating a convincing spear-phishing email to running a reverse shell. This is also achievable, due to the simplicity that the Platform offered through its utility, in which the user is able to supply a set of unstructured code even in plain-text English, and obtaining a ‘code set’ in-return, with the intended patterns of operations. The question at hand is: whether this is just a hypothetical threat or are there threat actors already using OpenAI technologies for malicious purposes. A deep-rooted investigation, into the web domain; shows that, there are several major underground hacking communities who are already using the available feature of OpenAI to develop malicious tools. “Cyber Secure India”, is also able to speculate, that in more than many cases among users of ChatGPT; there are a large number of cybercriminals using OpenAI with NO or NIL development skills. Also, we at “Cyber Secure India” believe, that it is a matter of time until more sophisticated threat actors enhance the way they use AI-based tools for Machiavellian actions.