Lab Manager | Run Your Lab Like a Business
Food scientist examines data on a tablet

Detecting Food Fraud through AI-Powered Data Analysis

AI and ML open the door for faster, more efficient food fraud detection by data analysts

by
Karl Ritter

Karl Ritter spent more than 30 years at Mars Inc. in the analytical, color, and flavor laboratories; the last eight years of which as the analytical lab manager for the...

ViewFull Profile.
Learn about ourEditorial Policies.
Register for free to listen to this article
Listen with Speechify
0:00
5:00

Seldom does a day go by without a news item telling us how AI will change our lives. The list of how artificial intelligence (AI) and machine learning (ML) impacts the human race is constantly growing. From self-driving cars to healthcare management approaches where ML algorithms help pathologists analyze tissue samples, to generative AI which can create text, imagery, and audio to meet the exact needs of the client. Just the very act of asking Siri or Alexa a question shows how AI has permeated so many aspects of our daily life. While there are debates about the pros and cons of this technology, one area that will prove to be a great help to humankind is the use of AI/ML in detecting food fraud and improving food safety.

AI is the ability of a computer to perform statistical analysis on large sets of data and use that information to analyze and learn. Algorithms are developed from this learning and can be applied to specific needs. This learning is an automated process and allows for continuous learning to occur to constantly improve these algorithms. ML is a particular type of AI where machines learn how to better respond based on structured big data sets and ongoing feedback from humans and algorithms.

Get training in Chemical Hygiene and earn CEUs.One of over 25 IACET-accredited courses in the Academy.
Chemical Hygiene Course

AI/ML is actively being applied to one of the major issues affecting the world today: food fraud. Food fraud is the deliberate attempt by a food supplier to deceive the consumer about the quality, origin, or composition of a food item. According to the US Food and Drug Administration, it is estimated that the cost of food fraud to the food industry is around $10-15 billion per year with some estimates going as high as $40 billion per year.1 In addition to the monetary impact, consumer safety and confidence are at risk. Food fraud stretches across many different parts of the food industry—just about every food group (beverages, vegetable oils, seafood, meat/poultry, etc.) have examples where fraud occurs on a regular basis.

While fear about AI has been growing recently, it is important to note the ways in which it can benefit our lives.

Historically, the food industry has taken a multi-pronged approach to preventing food fraud. By performing raw material risk assessments, high-risk supply chains and materials are identified. Based upon those assessments, analytical testing occurs to look for known adulterants in specific foods to identify fraud. This targeted approach works well when adulterants are known and the food fraud approaches are static. However, food fraud is an incredibly dynamic area. Fraudsters are constantly looking for new ways and new adulterants to fool analytical chemists. This sets up an ongoing battle between the fraudsters and the scientists, as well as reinforcing the need for a non-targeted approach.

Leveraging AI in detecting food fraud

This is where AI comes into play. With its ability to identify trends in massive amounts of data, AI can be used to identify fraud that would be undetectable using previous approaches. Research papers and programs combining chemistry and AI can be found across the web.

One such program is the US National Institute of Standards and Technology’s (NIST’s) Machine Learning to Predict Food Provenance. Food provenance is defined as where the food originated. Location can have a significant impact on food quality and consumer perception. This results in a significant impact on the economy since adulterated materials can be sold as higher quality products. At the NIST Food Safety Workshop in 2019, NIST identified the creation of computational tools and databases to determine food authenticity as one of the four main pillars of Global Food Safety. They stated that the mathematical techniques used to analyze large datasets of chemical and biological data for food authenticity is lacking. The goal of the NIST project is to develop chemometric fingerprints to ensure the safety and security of the US food supply. By partnering with other agencies to review existing databases, they work to determine the plausibility of using chemometric signatures as predictors of authenticity and provenance. These AI/ML approaches will allow for increased and more rapid testing to occur. 

Literature abounds on the use of AI/ML in food fraud detection. Just a few examples include A. Aznan et al. (2022) using near-infrared spectroscopy and E-nose sensors to detect fraudulent rice samples where the origins of the rice were in question or premium rice was diluted with lower-quality rice.² Similarly, T. Chung et al. (2022) reported on the use of Fourier-transform infrared spectroscopy data with a ML model to examine 18 different adulterants added to milk, ranging from melamine to water.³ B. Mithun et al. (2018) used AI/ML to evaluate hyperspectral sensing and RGB imaging data to detect the use of a carcinogen (calcium carbide) as a ripening agent in artificially ripened bananas.4 These reports, along with many more, show how AI/ML is being used to conduct a non-targeted approach to use analytical data in ways never done before. 

How data analysts and chemists can work together

This approach can be foreign to analytical chemists from yesteryear. Previously, to detect an adulterant, you would need to know what the adulterant was and then develop an analytical method for testing for that adulterant. For years, this targeted approach was the main approach. But computers evolved, and now it’s possible to efficiently analyze large sets of data. It is now the data scientist who is looking at the results to identify when a sample looks different than normal and is therefore adulterated. You no longer need to know what compound you are looking for, but rather just determine if the material is the same or different. This non-targeted approach is where AI and ML truly shine. The data sets can be from any analytical technique—mass spectrometry, UV-Vis spectroscopy, nuclear magnetic resonance, infrared spectroscopy, direct scanning calorimetry, etc.

With its ability to identify trends in massive amounts of data, AI can be used to identify fraud that would be undetectable using previous approaches.

One item that must always be considered, however, is the classical computer science concept of garbage in, garbage out. If the analytical testing is not done to perfection, the data being analyzed by the AI algorithms will be flawed and inaccurate conclusions will be made. Therefore, chemists’ skills remain a valuable part of the equation. Along those lines, new degrees are being developed to combine the skill sets of chemists and data scientists. The University of Washington now offers a PhD in chemistry with a data science option. Similarly, Northeastern University offers a bachelor’s degree in data science and chemistry, which is actually housed in the Khoury College of Computer Sciences. This merging of sciences is key for chemistry-based AI to reach its full potential. 

While fear about AI has been growing recently, it is important to note the ways in which it can benefit our lives. In order to maximize AI’s potential in detecting food fraud, new and robust chemical databases will need to be developed and shared. The secretive world of food companies will need to open up and share data so that robust algorithms can be developed and used to provide a safe and reliable food chain.

References:

1.    "Economically Motivated Adulteration (Food Fraud)." https://www.fda.gov/food/compliance-enforcement-food/economically-motivated-adulteration-food-fraud.

2.    Aznan, A.; Gonzalez Viejo, C.; Pang, A.; Fuentes, S. "Rapid Detection of Fraudulent Rice Using Low-Cost Digital Sensing Devices and Machine Learning." Sensors 2022, 22, 8655. https://doi.org/10.3390/s22228655.

3.    Chung, T. et. al. "Non-targeted detection of food adulteration using an ensemble machine-learning model." Scientific Reports 2022, 12, 20956. https://doi.org/10.1038/s41598-022-25452-3.

4.    Mithun, B.S.; Sujit Shinde; Karan Bhavsar, Arijit Chowdhury, Shalini Mukhopadhyay, Kavya Gupta, Brojeshwar Bhowmick, and Sanjay Kimbahune. "Non-destructive method to detect artificially ripened banana using hyperspectral sensing and RGB imaging." Proc. SPIE 10665, Sensing for Agriculture and Food Quality and Safety X, 106650T (15 May 2018). https://doi.org/10.1117/12.2306367.