Exponent Faces^TM

API Forensics^TM is proud to announce the release of Exponent Faces^TM, the next product in a series of offerings belonging to the Exponent^TM library of DFIR add-on X-Tensions for X-Ways Forensics (XWF).

Exponent Faces^TM implements leading-edge commercial facial recognition technology (in use by military and police organizations), directly within X-Ways Forensics, to detect, match and extract faces from photographs and video files. Forensic examiners can quickly and accurately identify victims, missing person and persons of interest within volumes of collected media, including security surveillance video.

After a long road of development, we are excited to be able to bring this powerful new tool to law enforcement that will allow forensic investigators to cut through their media-heavy case backlogs.

 

As a former police investigator, I know how daunting it is to sit and systematically comb through thousands of videos and photographs that might contain key evidence that can’t be overlooked. Trying to identify and rescue potential victims or missing persons or identify persons of interest requires investigators to physically review these thousands of images, or hours upon hours of video footage, in real time.

 

While this is not the best time effective use of police resources, this level or degree of thoroughness is what it takes to make sure no stone is left unturned. The casesloads are only getting bigger and at API Forensics, we recognized that.

 

When it comes to video files, sometimes the only clue to solving a case rests will evidence discoverable only at a fraction of one second, between frames. To the naked eye, this could be easily overlooked. Unfortunately, the reality is that investigators may not be able to go through every single video (or photograph). This could be for a number of reasons. It could be due to a sheer volume of evidence, not enough human resources, or simply not enough time, often attributed to competing case priorities or investigative urgency. This is an unfortunate reality and one that can lead to many risks.

 

We get it! We understand. Law enforcement agencies are doing the best they can and a lot of the times, they wish they could do more. The technology simply has not existed and/or there simply hasn’t been enough time. Today however, we’re hoping that can change for investigators. Exponent Faces^TM has the potential to make a big dent for many forensic practitioners.

 

In a lot of cases, police investigators want to detect and identify new persons, in addition to those who might already have been identified as victims or persons of interest through other courses of their investigation. This can be a time-sensitive and often mission-critical objective in a media based investigation. Having a tool like Exponent Faces^TM could change the tide when it comes to cases involving CSAM where detection and protection of victims is paramount.

 

Going through all of the evidence, or at least a good portion of it, no longer has to be a painful and time consuming task.

– John Bradley
Founder & CEO

Using Exponent Faces^TM – BEFORE

Using Exponent Faces^TM – AFTER

About The Science Behind It

The effective outcomes of the use of facial recognition technology are determined by a number of impact factors, which include but are not limited to:

• The quality of the image itself and the resolution
• The angle, position and minimal obstructions of persons’ faces
• The vertical and horizontal rotation of the head (measure in degrees °)
• For videos, the speed of the footage measured in frames per second
• The science behind face detection and confidence of similarity algorithms for matching

Exponent Faces^TM implements a top tier commercial facial recognition technology that has been in use for many years. In particular, this technology is in use today by military and police agencies and security organizations, both domestic and international. The choice to adopt a reputable and scientifically backed and supported technology was crucial in withstanding judicial scrutiny and valuation of evidence. Our solution has been developed over the past few years with careful attention to how police work with media-based evidence.

How It Works

When it comes to videos, we recognize that this is where police will have the hardest time reviewing evidence. It also happens to be one of the most valuable and probative sources of evidence in any major criminal investigation. At API Forensics, our development team have been in their shoes. What’s important is to be able to slice a video down to its smallest part, capture it and preserve it in a forensic manner that will (a) aid investigators reviewing the results, and (b) make it easier to defend in court and disclose to stakeholders.

Users of the software have ability to extract frames at the smallest interval in milliseconds. They can then scan the frames to detect and extract faces. And finally, they can then search those faces for known persons that they enroll in the program. The really nice part about all of this is that the heavy lifting is done automatically for investigators.

Frames are automatically added to the X-Ways Forensics case as child objects to the parent file (e.g., the video file). Faces that are extracted, are then add to the same case, as as child objects of the frame objects. The following are sample naming conventions of both frames and faces and these are worth noting, primarily because they adopt forensic best practices and provide accurate chain of custody of the smallest parts:

• [ItemID = 1, ParentID = 0] MySampleVideo.mp4
• [ItemID = 2, ParentID = 1] MySampleVideo_00230000ms_FRAME.jpg
• [ItemID = 3, ParentID = 2] MySampleVideo_00230000ms_02_FACE.png
  (where ’02’ is the second face in the frame at offset 230,000ms or 230 seconds).

At the time of this release, we’ve also added the ability to extract segments of a video file that we call CLIPS. This is an exciting add-on piece of functionality that will certainly have many applications for investigators. Now it’s possbile to create CLIPs of small moments in time and add them to the case file as their own individual pieces of evidence. Exponent Face^TM makes it possible to extract clips from videos, then extract frames from clips, and then faces from frames — all in one pass!

Gender Recognition

Exponent Faces^TM supports gender recognition for both male and female with a high confidence rate. This newfound functionality makes it possible to filter detected faces into two distinguishable groups, which could be very significant to an investigation where gender is a key factor.

The gender for each face is recorded in X-Ways Forensics using the Labels: MALE or FEMALE.

Face Matching

The following video illustrate the capabilities of Exponent Faces^TM in being able to not only extract faces from photographic evidence, but to then be able to match them against known faces. This process uses the Face Enrollment feature which allows investigators to introduce faces that they are looking for, quite easily.

Extract Frames and Faces From Videos

Forensic examination of multimedia requires a lot of attention to detail, time, and more importantly, the right tools to view and analyze the evidence. When it comes to videos, there is a common set of problems when trying review these types of files.

With the modern day smartphone, videos are recorded in high quality at a rate of no less than 24 frames per second. That is equivalent to 41.66 millisecond intervals. Some devices record at 30 frames per second which translates to 33.33 millisecond intervals. Why do these values matter? It’s simple. It’s not uncommon for investigators to use a simple video player such a Windows Media Player to review these kinds of media files. But this comes with inherent limitations.

The first big problem with this approach is that during playback, if something catches the eye of the investigator, it is near impossible to pause the video in anything less than one-second intervals. Often, it’s more like 2 or 3 second intervals. While there might be more advanced players available out there that might provide the ability to step through in smaller increments, the best average playback rate of most players is 1 second.

Many forensic tools do not natively provide video playback capabilities, but instead spawn an instance of the default player on the user’s computer system.

Attempting to zero-in on video content with any critical level of detail, using the above approaches is simply futile. When you consider the speed of persons and objects being filmed, the trouble becomes evident when you contemplate the distance something can move inside of one second.

 

Being able to slow things down to its smallest point in time can be absolutely crucial in an investigation. This is where Exponent Faces^TM can make a difference. Using the customizable features to extract frames at specific frame rates (by the millisecond), investigators can now pull out granular details that they might have otherwise been unable to obtain.

This next video illustrates how Exponent Faces^TM can be used to extract CLIPS (sections of footage) from video files. This can be very useful in isolating a key point of interest in a particular video, without having to playback or disclose the entire file. It is also possible to optionally extract frames and detect and match faces at the same time.

Video Playback Within X-Ways Forensics

We’ve already mentioned how frames can be extracted from video files. We’ve also explained that users have the options to specify the interval (in milliseconds) at which they would like frames extracted and added to their case, within X-Ways Forensics (‘X-Ways’). As an incidental cause and effect of providing the features of frame extraction, in conjunction with X-Ways’ very responsive user interface, it is now possible to emulate a movie player directly within X-Ways.

How is that possible you say? It’s actually quite simple. Using the Table view to navigate from one record (frame) to the next, the contents of each selected frame are displayed in the Preview pane below. By moving from one record to the next, it is possible to step through the frames. However, by HOLDING DOWN (keeping depressed) the UP or DOWN keys on the keyboard for any extended period of time, the pictures will load so quickly into the Preview tab that it will create the illusion of a slow motion video playback.

Keep in mind that this capability (to display images responsively in the Preview pane) has technically already existed in X-Ways Forensics. However, its likely not been as noticeable or appreciated until working with Exponent Faces^TM. See the difference between using a frame rate of 20 FPS (50ms) and 4 PFS (250ms) for video playback. Each has its merits and purpose in an investigation. See it in action, here for yourself.

Applications for Exponent Faces

The facial recognition technology and frame extraction features offered by Exponent Faces^TM collectively form an invaluable tool for forensic and security practitioners everywhere. From law enforcement to government and private security organizations, Exponent Faces^TM provides powerful and fast capabilities to detect, extract and match faces across a myriad of media source devices. Here are some everyday examples.

Download Your Free Trial

Get started today and realize the full potential of Exponent Faces^TM in your X-Ways Forensics investigations. Simply fill out the download request form and we’ll send you an Exponent^TM 30-day trial license by email.