Cyberbullying and Grooming Detection

This system mainly detects texts sent and received from social network and messaging apps. It also detects more generic text such as the text that appears on the screen when you are navigating through an app.

Severity Levels

Detections are classified in three distinct ‘severity’ levels: low, mid and high. 

Popular apps where this is active

• Instagram
• Messanger Lite
• WhatsApp
• Snapchat
• Tik Tok
• Google Chrome

Examples of detectable texts

Send any of these messages to someone on Instagram, Messenger Lite, etc. and Dug will pick it up.

What does it look like?

On the Dug menu, it can be switched on/off and detections can be reviewed

Dangerous messages detection menu: See that the detections made over which app and the severity level associated to each detection.

Dangerous message detection detail: This view shows a context of the detected text by providing three screenshots of where the detection was triggered.

Unsafe Search Detection

This system detects app searches that can be potentially dangerous for kids. There are 3 main categories a dangerous search can fall into:

• Dangerous Virals: viral trends and challenges circulating on social media that may endanger the kid or expose him/her to inappropriate content. 
• Dangerous Influencers: social media content creators that post adult, misinformative or provocative content. 
• Adult Titles: movies and series titles that are recommended only for adults or older teenagers.

Popular apps where this is active

Examples of detectable searches

What does it look like?

On the Dug menu you can find the unsafe search detection, where it can be switched on/off and detections can be reviewed

Unsafe search detection menu: Here you can see all three types of detection. Note that the viral detection includes an information icon because further details can be accessed by clicking it.

Viral unsafe search detection detail: Includes an interactive web search to provide the user with further info.

Explicit Music Detection

Detect when a child is consuming inappropriate music through audio and video apps. The system currently supports detection over the 3 most used audio-streaming apps on the tablet:

• Spotify*
• YouTube
• YouTube Music

• Detection of explicit music in Spotify requires the parent to configure a specific setting inside the app. See below for instructions on how to turn it on.

Detection Flow

The detection system can be activated from the settings menu of the Tablet or the Parental App. Dug will detect the song being played from each compatible app and classify it as Explicit or Not Explicit.

What makes a song "explicit"?

A song is deemed ‘explicit’ if it contains an explicit label as defined by Spotify. The criterion for a song being labeled as explicit may be one of the following:

• References sex, drugs or the occult
• Uses strong language such as curse words

Examples of detectable songs

FAQs

How do I enable Spotify detection?
On your tablet, open Spotify and go into the settings section located in the upper right corner. Then, on the settings list, enable the Device Broadcast Status toggle button. All set! 

Why doesn't the tablet block the song nor show a message to the kid?
We created this system to alert parents of the music their children are hearing, but not interfere with the content that the kid is consuming. 

What does it look like?

On the Dug menu you can find the explicit music detection, where it can be switched on/off and you can find information about detections.

Explicit music detection menu: Song title and artist are shown for each detection. By clicking on a certain detection item, info about the artist and album cover can be viewed.

Explicit music detection detail: An image of the listened song album is displayed along with a text indicating the ratio of explicit songs this artist has released.

Mood Detection

Child depression is a serious issue affecting a growing rate of children and teenagers throughout the world. With the Mood Detection system, SoyMomo tablets can notify parents when their kid searches something that can be associated with a depression risk factor. 

Risk Factors Identified

Depression risk factors are a set of activities, mood states, or situations that can potentially be related to a depressive pattern the kid might be experiencing. The table below gives us some information on the risk factors covered by this detector.

• For this iteration of the mood detector, we only consider messages sent by the user and not those that he or she is receiving.

Apps where this system is active

The mood detection system works over a pre-defined set of applications, which are listed below:

What does it look like?

On the Dug menu you can find mood detection, where you can switch it on/off and access detection details.

Mood detection menu: Each detection has a label with the most worrying risk factor and the text with the highest danger score that triggered the detection. Suicide Related detections receive a special warning given their severity

Mood detection detail: where screenshots contextualizing detaction are shown

When you click on any screenshot, a zoomed in view is displayed.

Inappropriate Images Detection

This system consists solely of the inappropriate images detector, which looks for obscene imagery of what is being displayed on the tablet screen. When active, screenshots of the tablet screen are taken at about 500 ms intervals and passed through the detection pipeline in search of something obscene. When a detection results positive, the app displaying the content is closed and a warning screen is shown to the user.

Types of images detected

The inappropriate content search space is divided into three classes:

• Porn
• Hentai
• Sexy

What does it look like?

On the Dug menu you can find image detection. From here it be switched on/off and you can find the detections.

Inappropriate image detection menu: See the capture time along with an image of the content being blocked for each detection.

There is a blocking screen displayed on the tabet right after Dug detects an image with inappropriate content.