Proliferating web technology has contributed to information warfare in recent conflicts. Artificial Intelligence (AI) can play a significant role in stemming disinformation campaigns, cyber-attacks, and informing diplomacy in the rapidly evolving situation in Ukraine. Snorkel AI is dedicated to supporting the National Security community and other enterprise organizations with state-of-the-art AI technology. We see this as our responsibility in the service of protecting democratic values around the world. Snorkel AI is offering Snorkel Flow at no charge to support federal efforts in response to this conflict. Reach out to firstname.lastname@example.org to learn more.
The use of technology in the current conflict creates a daunting scale of impact. Russia is currently engaging in a three-front information offensive against Ukraine.
- Disinformation: Over the past decade, Russia has developed and honed the largest and most comprehensive influence operation in human history. These capabilities are now used in a targeted, adversarial approach to reduce the morale of Ukrainian citizens, bolster Russian sympathizers in the Donbas region, and drum up support for the campaign amongst Russian citizens. Twitter, Facebook, and Telegram accounts can be generated automatically or bought by the thousands as disinformation signal generators.
- Cyberwarfare: Russia is employing sophisticated cyberwarfare operations ranging from using malware to cripple Ukrainian weapon systems to shutting down access to government and infrastructure websites. One cyber probe can scan an entire network of systems looking for a single unpatched entry point. Once found, these entry points can be exploited through a massive, growing list of known common vulnerabilities and exposures (CVEs) and currently unknown zero-day exploits.
- Kinetic warfare: This past week, Russia has expanded into kinetic operations, entering the theater with almost 4x the military assets of Ukraine. These kinetic offensives use automated signal collection and AI-enabled signals intelligence (SIGINT) to guide precision weaponry and specialized army units towards high-value targets.
The concurrent use of these three strategies is not a secret. It is Russia’s published playbook.
How AI can help counter information warfare
Responding effectively to the deluge of attacks is not manageable by human analysts alone. Fortunately, state-of-the-art AI approaches can exponentially scale the work of analysts and experts combating adversary efforts, and deploying these approaches rapidly is more possible than ever.AI can be used in many ways to respond quickly to the information warfare operations being waged against Ukraine. Examples include:
- Disinformation: Use natural language processing (NLP) to automatically identify disinformation, then take immediate action by publishing factual responses and removing adversarial social media accounts.
- Cybersecurity: Develop anomaly detection applications to scale analysts’ ability to monitor vast networks and rapidly respond to incidents.
- Kinetic warfare: Use information extraction to analyze signals and adversary communications, identify high-value information, and use it to guide diplomacy and decision-making.
Snorkel AI is here to help
At Snorkel AI, we believe in defending the principles of a democratic society. We’re committed to empowering our customers in National Security and infrastructures to accelerate their AI capabilities and gain a technological advantage against our adversaries. The Snorkel AI team has spent more than half a decade researching and developing programmatic labeling before building the Snorkel Flow platform. Snorkel Flow replaces manual labeling with a programmatic approach to intelligently auto-label thousands of data points in minutes, helping users accelerate their AI development by 10-100x. The platform can be deployed securely and rapidly to deliver value in days. We have proven solutions in:
- NLP: A Fortune 50 bank used NLP to similarly classify relevant information using a news analytics application. With programmatic labeling, they labeled their training data 45x faster than was possible with prior hand-labeled techniques. The end model was 25% more accurate than a previous black box vendor system.
- CyberSecurity: A major AI center of excellence within the U.S. Government built a network anomaly detection application for this purpose. They used an existing network attack ontology to programmatically label training data, and the resulting model achieved 88.1% accuracy.
- Information Extraction and Analysis: A Fortune 500 company used information extraction in an AI application to extract valuable clinical trial information from unstructured data. They trained models on 50k programmatically labeled documents and achieved 95.9% accuracy.
Snorkel AI has experience helping a diverse set of customer organizations successfully use Snorkel Flow to rapidly develop solutions for use cases just like these, and more. Please see more example case studies from commercial organizations.