TATA Advanced Systems Limited’s Post

An interesting blog that the ORNA team have put together this week and a subject that I’m seeing more of as we move to the end of the year and people are looking at what threats to expect to increase in 2025. Let me and the team know what you think and if you have real world experience of this scenario or more to add then please share 🙏 https://lnkd.in/eyEGb9qC #cyber #security #ai #threatvector #incidentmanagement #intelligence #tabletopexercise #awareness #sharingiscaring 

I can't wait to optimise the full power of Ai through precised and proper prompting. I wish to achieve a landmark in my research in history security and cybercrime. #aice #aiced2024 #alx_ai

Artificial or manipulated content containing human subjects – often known as “deepfakes” – is generated by sophisticated computational techniques. This technology is increasingly common and credible, posing a significant threat to the Department of Defense (DoD), especially as U.S. adversaries use deepfakes for deception, fraud, disinformation, and other malicious activities. In order to improve the efficacy of its forensics and intelligence aggregation functions, the DoD is looking to field detection and attribution solutions. Details on desired solutions here: https://lnkd.in/edAwvP39 Solutions requested by 6/27/2024. #cyber #AI #ML

Developing adversarial mini agents capable of operating on highly resource-limited devices, such as laptops. The agents will be able to conduct research into trending adversarial attacks, generate testing inputs based on found research and assess the resulting model actions. This is in the preliminary stages: the main agent framework is incomplete, and no fine-tuning or augmentation yet. However, it's running efficiently on my laptop which is a major preliminary step for building the mini agents :) Leveraging directML, onnx runtime and phi3 mini, pretty impressed so far. Stay tuned. #malware, #ai, #informationsecurity, #blueteam #reverseengineering #cyberdefense #cybercrime, #cyberthreatintelligence, #cyberdefense, #cyberwarfare #networksecurity #sec #security #tools #offensivesecurity, #redteam #innovation

risks AI: security: mitigation: genAI:

risks AI: security: mitigation: genAI:

Threats from cyber-crime are increasing and becoming more sophisticated, representing an issue for businesses in all sectors and sizes. While cyber-crime is a threat to organisations, we also benefit from being in a region that has significant strengths and experience in tackling cyber-crime. In our first online team meeting of 2025 we heard from Professor Daniel Prince, Co-Director of Security Lancaster who outlined the work of the North West Cyber Corridor initiative to help business counter this increasing threat, including the issue of online fraud driven by AI, and how AI might change the nature of cyber-attacks. The key takeaways he gave us, was to celebrate how good the North West region is in this sector, but also that there is still huge potential to be unlocked. https://lnkd.in/gEiDfXMQ

AI's impact on cybersecurity is undeniable, but not all innovations make sense. While some products integrate AI effectively, many lack practicality. Attackers are harnessing AI for dangerous purposes, particularly in phishing and deep fakes, introducing new threats. Enterprises vary in AI adoption, with some blocking it entirely while others explore its potential. Amidst the noise, it's imperative for organizations to scrutinize vendors' AI claims, prioritizing tangible outcomes over buzzwords. Learn more about AI's impact on cyber and fraud from our session at The Montgomery Summit presented by March Capital led by March partner Jed Leidheiser and featuring the top cybersecurity founders including Dave Merkel (Expel), Jon Miller (Halcyon), Noam Schwartz (ActiveFence), and Rinki Sethi (BILL). https://lnkd.in/drb7aqSB #montysummit #technology #innovation #cyber #cybersecurity #cyberthreats #AI #investing #founders

Cybercriminals are increasingly turning to deepfakes, a sophisticated form of digital deception that uses artificial intelligence to create fake but highly realistic videos, images, or audio. These manipulations can mislead audiences, damage reputations, spread misinformation, or commit fraud. From fake interviews to phony voice recordings, deepfakes have become a powerful tool for deception in both personal and professional contexts. What makes deepfakes particularly dangerous is their convincing nature. Unlike traditional photo edits or voice mimics, deepfakes are designed to look and sound authentic, making it hard for unsuspecting victims to detect the fraud. Cybercriminals use these AI-generated tools to manipulate public perception, scam individuals, or even blackmail their targets. As technology continues to advance, it's vital to stay informed about such emerging threats. Recognizing deepfakes and promoting digital literacy are crucial steps in safeguarding ourselves and our communities against this growing challenge. #Linkguard #Deepfakes #CyberSecurity #DigitalSafety #AIThreats #StayInformed

RAG Poisoning (PANDORA) – found something cool in the knowledge graph this am: “The integration of various plugins into LLMs, notably Retrieval Augmented Generation (RAG), which enables LLMs to incorporate external knowledge bases into their response generation such as GPTs, introduces new avenues for indirect jailbreak attacks. To fill this gap, we investigate indirect jailbreak attacks on LLMs, particularly GPTs, introducing a novel attack vector named Retrieval Augmented Generation Poisoning. This method, PANDORA, exploits the synergy between LLMs and RAG through prompt manipulation to generate unexpected responses. PANDORA uses maliciously crafted content to influence the RAG process, effectively initiating jailbreak attacks.” Paper: https://lnkd.in/eFXUwQtH #malware, #ai, #informationsecurity, #blueteam #reverseengineering #cyberdefense #cybercrime, #cyberthreatintelligence, #cyberdefense, #cyberwarfare #networksecurity #sec #security #tools #offensivesecurity, #redteam #innovation