In the past theArtificial intelligence has been abused to create malicious code and malware. This, however, does not mean that it can actually become a very effective cyber defense system.
A clear example of this is Gemini 1.5 Protool capable of processing up to 1 million tokens. Precisely this ability to store data makes Google’s AI ideal for analyzing malware and its evolution.
Exactly this LLMin fact, successfully identified its first zero-day vulnerability. The same had not been detected before by any type of antivirus, with Gemini 1.5 Pro using its deep knowledge of the code to identify traces of illicit malware activity. The malicious agent in question was apparently made to steal cryptocurrencies and be elusive compared to classic detection tools.
John SmithChief Researcher of the Gemini project, explained how “By analyzing the entire code at once, Gemini 1.5 Pro gains a total understanding of the malware, enabling a more accurate and complete analysis“.
Gemini 1.5 Pro cannot replace antiviruses yet, but it is a useful tool for cybersecurity experts
Gemini 1.5 Pro is trained with a large code dataset, including high-level languages, allowing the LLM to simulate the reasoning of a cybersecurity expert.
Not only that: AI has also been put to the test when it comes to i ransomware. This is what emerges from a test that saw Gemini analyze the binary files of WannaCrydemonstrating how AI has precisely identified both the characteristics of the malicious agent and the potential vectors exploited by cybercriminals.
While AI cannot yet completely replace specific tools, the ability to analyze behavioral patterns and anomalies can be of great help to traditional antiviruses. In fact, it should not be forgotten that cybercrime is constantly evolving and blindly relying on one tool could have disastrous consequences.