bugs we have today is significantly lower, due in part to that. But security has to be top of mind.” We’ve got options Cybersecurity attacks are becoming increasingly sophisticated – and they are only going to get worse. AI will likely play a big part, driving stream-scraping, automated account takeovers and signal manipulation using deepfakes. The industry needs a coordinated defence strategy that prevents these breaches altogether. “Looking ahead, the biggest challenge is in recognising that
as possible – and there are no guarantees anymore that that’s going to be the case,” says Corley. That’s why content provenance and authenticity are essential; provenance records a video’s origin, history and movements, while authenticity verifies that it hasn’t been altered artificially. By moderating at the moment of ingest – rather than after distribution – organisations can reject manipulated content in real time. Wowza’s latest partnership with NVIDIA allows for exactly this; the company’s Video Intelligence Framework (VIF) now includes NVIDIA’s Synthetic Video Detector, so its customers can flag AI-generated video in-stream, before it goes live and does any damage. “Once you purchase a VIF licence, it will take a stream and start to run inference on it with a model that’s provided by the customer,” Corley explains. “The customer, more or less, would train that model to detect scenes or objects in a stream that they’re looking for.” With the synthetic video detector, it works in largely the same way. “That model, in real time, is able to detect whether or not a stream is real.” According to a report by Surfshark, synthetic video has caused major financial losses in the media and broadcast sector, with impersonations of public figures driving a net loss of $400m alone. In other industries, AI-generated videos have been used to trigger unnecessary emergency responses, fabricate crimes and interfere with biometrics – with consequences far beyond simply being misinformed. “The pressure is coming down on everybody,” admits Corley. “We all face a common enemy at this point: malware and misintent. AI definitely increases that risk, and staying on top of that is becoming more and more key.” Ironically, AI is being used to improve cybersecurity too. “AI can hurt you and help you,” Corley states. “You can’t just connect AI to anything and let it rip. I could tell AI to go and look at my code and find vulnerabilities before they ever become released. The number of
» Cybersecurity attacks are becoming increasingly sophisticated – and are only going to get worse. AI will likely play a big part «
content security and session security have become equally important,” argues Hynard. Approaches like Castlabs’ continuous token authentication, coupled with multi- DRM deployments, are one way to protect video at the point of playout. Another option is to apply zero- trust architecture to the entire video pipeline, verifying every frame, every token and never trusting a given account, device or IP address. It is certainly a strict, cynical approach and one that potentially decreases productivity, but it mitigates both internal and external threats and better protects remote and hybrid workflows. Finally, forensic watermarking, while reactive, ensures that any video pirate will be caught red- handed. This is especially useful for high-value media such as live sports, big-budget films and flagship TV series – particularly those that haven’t yet been released. This protects creators from copyright infringement and ensures they are paid their fair share – because while a security breach might sound bad, no one wants to be short-changed.
Powered by FlippingBook