Detect COVID-19 Phishing with Machine Learning

Covid-19 has had a major effect on the world. With millions of people told to self-isolate, the working day for many has been turned on its head. Remote working is now in full swing and companies are running officeless. Whereas this is great for productivity, it increases the danger that cybercriminals pose to companies. With more and more communication being done through email, the opportunity for phishing attacks has never been higher.

Here at Red Sift, we’ve built a machine learning model to distinguish fake COVID-19 messages from the true ones based on the email content. This blog post explains why we took a Machine Learning approach and how we built the model.

Why use Machine Learning?

A traditional rule-based approach to classify email intent isn’t scalable because a large number of rules need to be manually crafted and maintained over time. Attackers are smart and they will constantly find a way to adapt their phishing strategy. For example, to get past a ‘covid-19’ keyword filter, attackers can simply change the keyword to a lookalike one such as ‘côvíd-19’.

Machine Learning uses mathematics and statistics to learn patterns from the data and make educated predictions about unseen data. Machine Learning models need labeled data instead of hand-crafted rules. Machine Learning algorithms can train a model to learn sophisticated and subtle patterns from example phish emails. The model can be easily updated to be more accurate with additional training data collected over time. As a result, an ML-based approach is more robust and dynamic.

Real life examples

This section shows some real-world examples of true and fake COVID-19 messages. Our model correctly classifies them all.

Example 1

Fake WHO message with malicious attachment using accent characters côvíd and misspelling word coronaaviris within the document.

Example 2

Fake government message with malicious links.

A legitimate Covid-19 email

A true COVID message from an organisation.

How we built the model

There are two important building blocks of a machine learning model: data and algorithm.


To effectively train a model, we need to provide a good collection of examples for both fake and true COVID-19 related messages. For true messages, we crowd-source hundreds of messages sent by a large number of different organizations. For fake messages, we filter them from reliable reported phishing messages. To increase variety, we collect fake messages in image format in a number of articles and extract text from them. Luckily, part of working at a cybersecurity company means that there was no shortage of scam messages!


We investigate a number of preprocessing techniques together with different modelling approaches. The preprocessing techniques include both generic text (stop words removal, lemmatization) and email specific (signature exclusion). The modelling techniques range from standard algorithms (logistic regression, naive Bayes, SVM, random forest) to shallow neural networks (word2vec, fasttext) and deep learning (RNN, LSTM). In particular, we make use of the state-of-the-art research in transfer learning including ULMFiT, transformer and BERT.

All this work means that we are able to give you 360° Email Protection at this tricky time, with OnDMARC preventing email impersonation, OnINBOX protecting you from inbound email threats, and OnDOMAIN securing your domain perimeter.

Red Sift find out more


Phong Nguyen

9 Apr. 2020



Recent Posts


Introducing DNS Guardian: Stop impersonation and spam caused by domain takeovers 

Rahul Powar

tl;dr: We’re thrilled to announce DNS Guardian — a new feature in Red Sift OnDMARC that can swiftly identify and stop domain takeovers that lead to malicious mail. Back in February, we shared updates with the community about SubdoMailing – an attack discovered by Guardio Labs. The attack was a form of subdomain takeover,…

Read more

Navigating the Information Security Landscape: ISO 27001 vs. SOC 2

Red Sift

As cyber threats evolve, so do the standards and frameworks designed to combat them. Two of the most recognized standards in information security are ISO 27001 and SOC 2. What sets them apart, and which one is right for your organization? Let’s delve into the key differences. Purpose and Scope: Global Framework vs. Client-Centric…

Read more

G2 Summer 2024 Report: Red Sift OnDMARC’s Winning Streak Continues

Francesca Rünger-Field

We’re delighted to announce that Red Sift OnDMARC has again been named a Leader in G2’s DMARC category for Summer 2024. This recognition is based on our high Customer Satisfaction scores and strong market presence. Red Sift appeared in 11 reports – 5 new ones since Spring 2024! – earning 5 badges: A few…

Read more

Google will no longer trust Entrust certificates from October 2024

Red Sift

Tl;dr: Google has announced that as of October 31, 2024, Chrome will no longer trust certificates signed by Entrust root certificates. While there is no immediate impact on existing certificates or those issued before 31st October 2024, organizations should start reviewing their estate now. On Thursday 27th June 2024, Google announced that it had…

Read more

Understanding the domain attack

Francesca Rünger-Field

tl;dr: The recent compromise of the domain has triggered a broad-reaching web supply chain attack, impacting over 100,000 websites across various sectors including finance, healthcare, non-profits, academia, and more. To ensure the security of your website, we strongly advise you immediately remove any reference to Latest update: 27th June 2024 Sansec, a…

Read more