What are Deepfakes: The term "deepfake" combines the deep learning concept of Artificial Intelligence (AI) with something fake. Deepfakes combine genuine images, audio, and video with AI to produce a convincing synthetic result.
Applications and Concerns: Although deepfakes can be used for positive applications, frequently, that result is used for nefarious purposes, creating people and situations that do not exist or did not happen to mislead groups of people into believing misinformation and intimidating them to compromise the safety, security, data of an organization, for example:
- Election manipulation: It is used to spread fake videos affecting election campaigns.
- Social Engineering: Deepfakes are used to create scams and get individuals to divulge sensitive information.
- Automated Disinformation Attacks: Deepfake can be used to spread disinformation attacks, such as conspiracy theories and incorrect theories about political and social issues.
- Identity Theft: deepfake technology can be used to create new identities and steal the identities of real people.
- Financial Fraud: Attackers use the technology to create false documents or fake their victim's voice, which enables them to create accounts, be helped by call center agents, or purchase products by pretending to be that person.
- Scams and Hoaxes: Deepfakes can emulate recognizable voices for robocalls to consumers and are often used in imposter scams that spread misinformation, endorse products, or steal money.
Legal and Ethical Challenges:
- Legislation: Federal bills and state laws still need to regulate deepfake usage.
- Data Privacy: Deepfakes raise red flags regarding privacy and consent.
Real-life Incident: In February of 2024, a group of fraudsters tricked a worker in Japan from the British engineering firm Arup into paying USD 25 million to fraudsters who utilized deepfake technology to imitate and recreate several other members of staff in a fake video call. The worker doubted the veracity of the CFO's initial request that came via email to make the transaction, even though he thought it was a phishing email. However, after attending the video call and seeing that the request was truthful, the worker put aside his concerns and proceeded with the financial transaction [1].
How do you identify if a video or image is deepfake-originated? Deepfakes can be spotted by recognizing unnatural movement or unusual activity, such as:
- Unnatural eye movement or lack of movement or lack of blinking
- Unnatural facial expressions (missing emotions) and facial morphing
- Unnatural hair
- Teeth looking unreal
- Abnormal skin colors or odd lighting or discoloration
- Awkward head and body positioning
- Inconsistent facial positions or improper lip-syncing
Recommendations to keep yourself safe from deepfakes:
- Be extremely cautious about what and how much personal information you share online.
- Limit the number of pictures and videos shared on social media, especially high-quality ones.
- Make sure to use private social media accounts.
- Consider using a digital watermark on images or videos published online. This can discourage deepfake creators from using such content.
To learn more about deepfakes, tune into this SHRM podcast: The AI + HI Podcast -
