Security is a broad notion that pertains not only to providing a service to businesses and individuals in their quest for safety but also to protecting their personal information and data.
Security for businesses entails providing a safe and secure environment for their employees and valuable equipment while that for individuals relates to the freedom of conducting one's normal day activities without the worry of being insecure.
While a lot of investment (in the name of security) has been done for surveillance and monitoring, the same cannot be said when it comes to protecting, preserving and securing the basic rights of individuals i.e. right to privacy in their daily life activities.
With growing use of technology, we see greater and larger expansions of video monitoring services. Be it shopping at the mall, stopping at a traffic light, catching a flight at the airport, etc, people and their activities are constantly being monitored.
Although this is extremely important given the volatile world we live in, there needs to be more debate on how and why people's privacy is at risk and what needs to be done to address this issue.
Video surveillance cameras, such as those included in a CCTV system, are used as safety measures by documenting persons' behaviors. Individuals are exposed to a great extent through video monitoring in the name of safety or security.
The extensive deployment of surveillance cameras captures illegal behaviors like theft, assault, and shooting occurrences, which benefits most citizens.
Nonetheless, the emergence of greater CCTV-based security has additional issues in terms of protecting people's privacy, dignity, and free will, even when they are being observed. Privacy invasion has been extensively and repeatedly reported as a result of the widespread deployment of surveillance cameras.
Video redaction is a mechanism for helping protect people's privacy. This is a technique for obscuring a data subject's personally identifiable and sensitive information included inside a video. It is applied using various automated methods to maintain privacy when recording or storing videos.
In redacted videos, entire objects (individuals), faces, vehicles, license plates, backgrounds, and complete video frames can be distorted automatically to obscure the data subject's identity.
(1) Object detection/tracking
(2) Object obfuscation techniques
Manual redaction of video can be accomplished by detecting objects or Regions-of-Interest (ROI), or it can be performed automatically by detecting and tracking an item through the scene. Object detection can be performed manually (someone draws a box around all objects of interest). This works but it is a slow, manual and cumbersome process. As opposed to this, there are automatic methods (using Deep Learning) of finding objects in a video which are highly accurate and thus save a lot of time.
Computer Vision (CV) is a subfield of Artificial Intelligence (AI) and comprises any expert/intelligent system that analyzes visual data to identify specific objects or regions of interest.
When it comes to object detection, recognition and tracking, these computer vision algorithms are becoming increasingly powerful. Object detection identifies real-world items in images or videos, such as faces, bodies, bicycles, buildings, and license plates.
Traditionally, object identification algorithms used extracted features and learning techniques to identify certain item category occurrences. They then employ object categorization and localization algorithms to ensure the accuracy of their results.
Thus, to provide a safe Data protection-by-design solution for any CCTV system, CV utilizes either human ROI selection or automatic detection/tracking using pattern recognition and machine learning algorithms (including Deep learning).
Following that, visible privacy protection mechanisms such as video redaction will be used.
ANONYMISATION VS PSEUDONYMISATION GDPR took effect in May 2018, requiring surveillance-data gathering organizations to recognize the keywords and practices specified by the regulation.
Anonymisation and pseudonymization are two distinct terms and must be grasped in order to provide GDPR-compliant safeguards for a subject's collected visual personal data.
This is the process of securing data through the use of an irreversible automated technique. The term irreversible is critical to grasp, as anonymous data cannot be restored to its original state or utilized to identify the data subject using any practical method.
This refers to the process of securing data by automatically replacing genuine personal information with a pseudonym so that the data subject cannot be directly accessed.
Although pseudonymization has a variety of applications, it should be separated from anonymization since it often provides only a limited level of protection for data subjects' identities, enabling indirect identification.
Where a pseudonym is utilized, identifying the data subject is often achievable through analysis of the underlying or associated data. In some instances, the data may be reversible/identifiable.
The term "reversible" here refers to the fact that the data subject can be identified using any other relevant information or automated mechanism, even if the data is not transformed back to its original form.
Thus, one may distinguish the two words by noting that pseudonymization is a reversible safeguarding technique, but anonymization is an irreversible safeguarding technique. Irreversible anonymized data is exempt from the scope of regulation, and controllers may retain this type of data indefinitely (for future statistical analysis) without violating the terms of the GDPR.
Controllers cannot return data to its original state for any purpose, even if the processing is legal.
Thus, it may be deemed a waste of storage space on the CCTV controller's end to save this type of data intended for future statistical analysis, assuming that processing is legal at some distant, future period, even if it is not presently permissible.
Image/video processing is often utilized to offer visual security for CCTV surveillance footage. Image processing is the technique of manipulating a digital image to enhance (or deteriorate) its visual quality. It entails a variety of manipulations.
For instance, image segmentation is used to identify pixel color images; geometric transformations (enlargement, reduction, and rotation) are performed; images are combined or blended; (4) image editing is performed, and (5) interpolation is performed.
It's worth noting that irreversible schemes can potentially be reverted to their original state or identified/recognized again only provided the originals are stored for future access.
Most AI-based image and video redaction tools accessible over the internet store the originals in case the modified image or video has to be recovered. The most widely used video redaction techniques are classified.
Only some known image and video redaction techniques are classified as Irreversible Video Redaction Techniques.
These include the following:
This form of redaction is accomplished through the use of image filters. The blurring filters can be applied to the whole video frame or a single region/object, such as a face, person, license plate, or sign.
The three most frequently used blurring filters are the mean filter, the weighted average filter, and the Gaussian filter. The Gaussian filter is used in most privacy-protection applications to provide effective blurring.
Pixelation is the method of expanding pixels inside images to create a blurred effect. Additionally, it is conducted via pixel interpolation to achieve strong distortion effects. Interpolation operates in two directions by estimating the values at unknown sites using known data points.
Thus, picture interpolation happens when an image is distorted from one pixel grid to another.
Although pixelation currently provides an irreversible redaction technique, researchers are developing an AI technology to reverse the pixelation effect and identify the faces hidden behind blurred images.
Since this technology may be reversed in the near future, using Sighthound Redactor's Mosaic Redaction Technique is the preferable alternative.
Small blocks of pixels from various regions are mixed to create an unidentifiable effect in the image; for example, pixel blocks cover a face from various sections of the image.
Video redaction can be used for various purposes, including protecting an individual's privacy, a business, or an organization. For instance, when a problematic situation involving the police department arises, the public often demands complete clarity.
However, before the footage is made available at the public's request, the police agency must adhere to local, state, and federal privacy laws. Before a video is released, it is necessary to redact information from the images, such as faces, registration plates, or other personally-identifying information.
Certain organizations may believe otherwise or be burdened by redaction expenses. Redacting a brief 20-minute video manually might take hours, but there are ways to reduce the cost and the time spent on redaction.
As a software service, redaction and video editing solutions can be affordable. Sighthound Redactor is the only redaction software that provides the Mosaic Redaction Technique.
Redaction software should be simple for both novice and expert software users. The simplicity of Sighthound Redactor software offers quick and effective redaction that incorporates artificial intelligence capable of tracking faces or other things automatically.
Sighthound Redactor offers quality redaction software that includes intelligent features, intelligent algorithms, an intuitive design and a user interface that guarantees a more accurate and quicker redaction process.