By Vineet Bahal
Deepfake emerges through the application of artificial intelligence and deep learning algorithms, generating remarkably realistic and deceptive multimedia content by overlaying or synthesising facial expressions, gestures, and voices onto pre-existing images or videos. Top of Form
However, as technology and its availability advance, the potential for misuse grows, and malicious usage is more and more likely to proliferate, impacting individuals as well as corporates.
The Rapid Evolution of Deepfake Technology
One of the most alarming aspects of deepfake technology is how rapidly it is evolving. Researchers have developed AI tools that allow users to edit video transcripts to alter words from someone’s mouth. Although such technology is not as readily available to consumers yet, it highlights the ease with which deepfake manipulation can be achieved. This fast-paced evolution of deepfake technology is a growing concern for individuals and organisations.
Generative AI and Deepfakes
Generative Artificial Intelligence (GenAI) techniques have been pivoting rapidly. A good illustration is the “Deep Video Portraits” system developed at Stanford University, which can manipulate facial expressions, eye gaze, 3D head positions, blinking, and head rotation through generative neural networks. Although the videos created may not be perfect, they look incredibly realistic, making it harder to differentiate between real and fake.
Challenging the idea that only those with comprehensive documentation can be used for deepfake manipulation, an AI lab has demonstrated “living portraits” of iconic figures. Machine learning can create these portraits smoothly, using just a few pictures, as it can now be done with general video footage of a wide range of people.
Such advancements have made it easier for everyday internet users to generate content by just using a basic prompt, thus providing an opportunity for the spread of cybercrime activities, including the growing use of deepfakes.
Detecting Deepfakes: Art and Science
Detecting deepfakes has become a challenging endeavour. In the past, online video and audio presentations were generally accepted as authentic. However, today, discerning what is real from what is fake can be a combination of art and science. Human intuition plays a vital role in detecting irregular vocal cadences, unrealistic shadows around the eyes, or other telltale signs of deepfake manipulation. More often than not, a deepfake may leave viewers with a lingering sense that something is off due to errors in the manipulation process.
There are several cues that individuals can watch for when trying to distinguish genuine content from deepfakes. These include incongruities in the skin and body, unrealistic skin tones, unnatural skin texture and gloss, unusual shadows around the eyes, atypical blinking patterns, unrealistic glare on eyeglasses, unnatural movements of the mouth, lip syncing, lip colouration discrepancies, facial hair inconsistencies, and unrealistic moles on the face. Staying vigilant and recognising these signs can be a crucial first step in early detection of deepfakes.
Creating Robust Security Procedures
With international business value chains and increasingly large amounts of enterprise data and sensitive content, information security is becoming a company’s most important pillar.
An organisation’s security strategy must include measures to prevent the harm caused by deepfake attacks. To alleviate the threat of deepfake attacks, companies may consult the Cybersecurity and Infrastructure Security Agency for guidance. The Zero Trust Maturity Model offers helpful advice in this regard.
Thwarting sophisticated deepfakes necessitates continuously monitoring and keeping up to date with the newest tools and technologies. Staying ahead of potential threats demands diligence and being well-informed.
Defence Against Deepfakes
Establish Deepfake Policies and Awareness (organisational perspective):It is essential first to establish comprehensive Deepfake Policies within your organisation to address detection, reporting, and responses to potential deepfake incidents. Educate staff about deepfake-related dangers and how to prevent them with top-notch best practices.
Protect Your Intellectual Property (IP):Implement robust cybersecurity measures to safeguard your IP and copyrighted content. Strengthen security protocols to prevent unauthorised access to sensitive materials.
Secure Storage and Transfer:It is important to utilise secure protocols for storing and transferring sensitive media to prevent data breaches. Implement encryption methods to enhance the security of your digital assets.
Watermark Sensitive Content:Embed watermarks in your sensitive content for easy identification and tracking, deterring potential manipulations.
Control Access to Social Media Profiles:Lock access to photos, media, and sensitive content on your social media profiles, enhancing privacy and control.
Utilise Deepfake Detection Tools:Employ advanced deepfake detection tools to verify the authenticity of media content. Cross-reference information with trusted sources to ensure the legitimacy of shared content.
Identify Anomalies in Content:Stay vigilant for anomalies, visual or aural artefacts, and irregularities in content that may indicate deepfake manipulation.
Check Media File Metadata:Scrutinise media file metadata for inconsistencies or anomalies in embedded information, providing clues to potential manipulations.
Verify Sources:Always check the source of provocative or remarkable content. Anything forwarded without proper attribution to the source is likely questionable; exercise caution.
By implementing these cybersecurity measures, you can fortify your defences against the rising threat of deepfake technology. Stay informed, stay vigilant.
The Future of Deepfake Attacks
Due to the swift progression of AI, predicting the future pattern of deepfake attacks is becoming incredibly difficult. The technology’s development is an ongoing process, which means deepfakes will likely continue to grow in prevalence and complexity. Consequently, it will become easier for malevolent actors to generate persuasive deepfakes and, thus, more challenging for detection.
Cooperation between public and private entities is paramount to counteract the widespread spread of deepfakes successfully. By doing so, we ensure secure online interactions and user trust and encourage honesty and sincerity in broader societal engagement. Despite the emergence of advanced deepfake detection technology, the threat of deepfake assaults is ever-evolving. Nevertheless, maintaining caution and flexibility is crucial to navigating this uncertain realm.
The author is COO, senior VP, delivery and operations, Nihilent Limited
Follow us on Twitter, Facebook, LinkedIn