51爆料网

Pope Francis in a puffer coat. Tom Cruise performing magic tricks. President Volodymyr Zelenskyy calling for surrender.

In the age of artificial intelligence (AI), seeing is no longer believing. From lighthearted memes and Instagram filters to phishing scams and deepfake propaganda, AI is blurring the line between fact and fiction鈥攁nd spreading falsehoods faster than truth can catch up.

鈥淭he things that fool us humans haven鈥檛 changed, but the ability to create those things at scale has really changed in the last five years,鈥� says Jason Davis, a research professor in the .

Through a Department of Defense grant, Davis and co-principal investigator Regina Luttrell鈥擭ewhouse senior associate dean and associate professor鈥攁re working with 51爆料网 Reserve Officers鈥� Training Corps (ROTC) Army cadets Glenn Miller 鈥�27, Daniel Canhedo 鈥�27 and Allison Simpson 鈥�27 to develop innovative tools that detect and defend against disinformation across text, image and video.

Deepfake Defense

Miller, a computer engineering major in the , is developing a deepfake detection model that verifies the authenticity of live video calls.

鈥淭he end goal is to have a model where if you鈥檙e on a Zoom call with someone, you can verify if that person is real or a deepfake,鈥� Miller explains. 鈥淚s this person actually who they say that they are?鈥�

The work builds on lessons learned from incidents like the Zelenskyy deepfake. 鈥淲ith the ever-evolving nature of AI and deepfake technology, it could be a very real threat of disinformation,鈥� Miller says.

Bot or Not?

Canhedo and Simpson, both computer science majors, are addressing another common challenge: distinguishing authentic online interactions from automated bots.

鈥淎 lot of YouTube comments are bots鈥攎any of them are phishing comments that try to get money from innocent people,鈥� Canhedo explains. 鈥淏asically, we鈥檙e working on creating a neural network machine learning model that can identify whether a YouTube comment is a bot or not.鈥�

Their 鈥淏ot or Not鈥� project involves building comprehensive datasets to train more accurate detection systems. 鈥淲e are gathering a lot of data,鈥� Simpson says. 鈥淲e need to collect both real comments and AI-generated comments to be able to train the machine learning models to be more accurate.鈥�

Newhouse Synthetic Media Lab?

Once students develop working models that have been repeatedly trained and tested using collected data, the models become tools in the newly launched . The online suite currently houses over 20 tools built, developed and continually updated by Syracuse student and faculty researchers to analyze text, images and video for signs of manipulation or deception.

鈥淭aking a really hard and seemingly difficult concept, compartmentalizing it into a series of steps, and then actually developing a finished product鈥攖hat was really rewarding,鈥� Canhedo says.

The tools in the Newhouse Synthetic Media Lab use a three-layer model to evaluate and respond to synthetic content:

1. Detection: Is it real or synthetic?
2. Attribution: Who made it? What tool or model was used?
3. Characterization: Is it malicious or benign? Factual or manipulative?

鈥淭hese detectors help us understand if we鈥檙e being fooled,鈥� Luttrell says. 鈥淲hen an image is synthetic, it changes how we analyze the content around it. That transparency alone plays a big role in helping us protect ourselves.鈥�

Preparing Future Leaders

With support from the Syracuse Office of Undergraduate Research and Creative Engagement, Davis and Luttrell have specifically sought ROTC students to work on their interdisciplinary research team spanning fields like communication, computer science and policy.

鈥淭hey鈥檙e just fantastic students,鈥� says Davis, who provides cadets with flexible remote research opportunities that align with their unique schedules and military commitments. 鈥淚 wanted to create opportunities for them to be able to do research that they find interesting, rewarding and that, most importantly, supports their next steps.鈥�

Nearly every one of their students has secured their first choice in deployment, including Christopher Nemeth 鈥�25, whom Miller, Canhedo and Simpson credit for getting them involved in their current projects. After working with Davis and Luttrell on the Defense Advanced Research Projects Agency鈥檚 Semantic Forensics project, Nemeth is now a second lieutenant in the U.S. Army鈥檚 Cyber Corps.

Canhedo and Simpson both hope to follow suit. 鈥淭hrough our projects, we鈥檝e been able to get a better understanding of the good and bad side of AI,鈥� Canhedo explains. 鈥淣ot only have we developed our technical understanding of things but also our ability to think more ethically and see how this can impact people. Overall, it is really awesome to be taking part in an initiative that鈥檚 so important and will only continue to grow in coming years.鈥�

鈥淭he research has definitely opened doors and given me a lot of experience working with other people鈥攚hich is especially important with where we鈥檙e headed as officers in the military,鈥� Miller says.

As AI-generated content becomes increasingly sophisticated, the work of Davis, Luttrell and their team offers a path forward. Their tools empower people to navigate the digital world with greater confidence and clarity. For these student researchers, the work represents an opportunity to shape a more transparent and trustworthy digital future.