Benefits for the security of citizens & selected tools and applications
Intelligent Monitoring for Threat Detection (Video Analytics)
INDECT applies a new approach to CCTV monitoring – known as Threat Monitoring. It focuses on the detection of scenarios containing potential threats. These scenarios include the presence of dangerous objects (e.g. guns), sounds (gun shots, screaming), abandoned luggage, etc.
Automatic recognition of sounds related to threats
Software is being developed that automatically detects sound events such as broken glass, gun shots, explosions, screams, and cries for help, and also determines the sources of such sounds. High accuracy of the systems will enable practical applications.
Potential application: Automatic detection and localisation of a threat based on sound analysis.
Detection of dangerous objects
Research groups are developing algorithms that are designed to detect dangerous objects in a person’s hands (e.g. a knife or a gun) in video footage. It is obvious that situations with a high degree of complexity (such as a crowded scene) may lead to false positives. Accuracy will be affected by such factors as the number and placement of cameras, and lighting conditions. Therefore, full analysis of the benefits may require live trials.
Detection of abandoned luggage
INDECT research groups are developing a system employing innovative video analytics algorithms for the automatic detection of abandoned luggage in airport terminals and railway stations. If the person who has abandoned an object does not return within a pre-defined time-period, a message can be sent warning of an unusual event.
Potential application: Automatic detection of abandoned luggage employing innovative video analytics algorithms on CCTV footage.
Detection of robberies and similar threats
INDECT is designing a tool which makes it possible to identify such events as robberies, including detection of human cries for help.
Potential application: Automatic detection of robberies based on analysis of CCTV recording.
System for threat monitoring based on social support network
The system is designed to monitor threats integrated with information coming from society and a Geographic Information System (GIS). The main focus is on improving distributed knowledge acquisition and management. An example scenario illustrates how citizens and police officers can cooperate with high efficiency and at low cost.
An example of the system would be the scenario where a citizen sends information about a traffic accident. Compared to existing emergency systems (such as “999”), the proposed system will provide more complete information (including the exact location of the accident, multimedia content, etc.), resulting in faster and more efficient response from the appropriate emergency services.
Potential application: Collecting information about threats from society.
Standardization of video sequence quality for user applications
Storage and transmission of video sequences have many applications outside the entertainment sector – this type of video sequence is usually adopted in task-based applications, including public safety, including video surveillance, medical services, remote command and control, and sign language.
Efforts to develop storage and transmission techniques are showing the great potential of digital platforms in accessing video surveillance recordings or video sequences with medical content. Video compression and transmission, however, cause the majority of common problems during the construction and use of these platforms. Video surveillance and medicine require a new dimension in quality requirements, as lossy compression techniques for video resources must be both efficient and reliable from the viewpoint of public safety and medical services practitioners.
Video service characterized by blurred images may result in serious consequences for a police officer or doctor. Therefore, the aim of the INDECT project is to develop the quality assessment of video sequences in user applications by cooperating in the drawing up of research methods (including any amendments to the International Telecommunications Union (ITU) specifications and standards) such as predictive models based on appropriate network parameters.
Urban photograph localization using the INSTREET application
INSTREET is a solution to the problem of geolocation of photographs by using an algorithm to compare their content against a geolocated database of street view images. The algorithm allows to pinpoint the location where a photograph was taken using visual similarity only in urban areas as hundreds of square kilometers. Potential applications of the tool can refer to tourist services, public safety, social media and search engines.
 Michał Grega, Seweryn Łach, Urban photograph localization using the INSTREET application—accuracy and performance analysis, Multimedia Tools and Applications
Threat detection in computer networks
INACT – Intelligent Advanced Image Catalogue Tool
The INACT computer forensic software tool is used for indexing and searching multimedia computer files. In the future, this tool could be used to help in the fight against crime related to the production, possession and distribution of child pornography.
INACT prepares evidence on the content of a directory where child pornography is suspected. INACT also automatically searches for suspicious files in the available resources. This allows rapid checking of the resource files for the presence of content identical or similar to that contained in the catalogue. The ability to find similar (not necessarily identical) files is one of the top applications of this software.
The application the project has created significantly expands the capabilities of existing investigative tools, and includes the following features:
- Build on the visual similarity between images to enable the detection of other images which have been resized, edited, or come from a photo session.
- Implementation of advanced automation mechanisms allows for approximately ten-fold speed increase in the search for a random search.
- The application has been created as the result of long-term co-operation with end users, which obviously enhances its usefulness.
Also being created currently is a complementary algorithm for the automatic detection of child pornography. The original algorithm is able to determine if a photo contains child pornography. The algorithm implemented does not assume the use of examples (it is not based on Query By Example methods), and this is a significant advantage.
With regard to ethical issues and privacy protection it should be stressed that:
- the INACT software is created with the aim of investigating the crime of producing child pornography, which is generally perceived as an unquestionably morally reprehensible.
- INACT can only be used to access the suspect's file repository. Current legislation requires a warrant which strictly defines rules for accessing such repositories.
More information can be found at:
 Michał Grega, Damian Bryk, Maciej Napora, Marcin Gusta, INACT - INDECT Advanced Image Cataloguing Tool. Multimedia Communications, Services and Security Communications in Computer and Information Science". 149, s. 28-36, 2011.
Potential application: Searching file repositories for illegal content (e.g. child pornography).
A Platform for Malware Network Analysis and Confinement
A botnet is a network of zombie computers compromised by malware (virus, worm). Botnets are coordinated by a bot-master through a command and control channel (C&C) to which the malware connects to get instructions. A bot-master can use botnets to perform malicious activities.
In this research, the architecture of a platform for studying botnets is being investigated, that includes finding adequate analysis methods, monitoring botnet activity, and finding efficient countermeasures or methods of confinement. The platform comprises a filtering and monitoring gateway, low and high interaction honeypots, and a system for malware analysis.
Potential application: Fighting cybercrime related to the use of malware.
Data and privacy protection
An important part of the project is research on ensuring information security and personal privacy. Digital watermarking and cryptographic algorithms are the main mechanisms used for this purpose.
Application of digital watermarking for privacy protection – secure access to information included in digital images
Digital images often contain content access to which should be strictly controlled. Examples include images of the face, the whole body or license plate numbers of vehicles. At the same time, these elements are frequently only a small portion of the photo, while the remainder of the image may still provide valuable information to a wider audience. This scenario is reflected in the case of applications such as Google StreetView, or streams from the traffic cameras, often published on the Internet.
In these applications, a key role is played by personal privacy. Therefore, techniques for detecting the presence of sensitive parts of the images and then blurring them are commonly used. This technique leads to an irreversible loss of information that is potentially useful in the future. Currently, if there is a need to preserve these sensitive data, two versions of the image, with and without censorship, are recorded at the same time. This greatly increases the chance of unauthorized people accessing this information.
The research conducted developed a technique to solve this problem. It enables sensitive areas of images to be blurred while leaving the "lost" information securely encoded directly in the same picture. This information is encoded using digital watermark technology, i.e. invisible modification of images that can be read by specialized software. A special decoder is able to reconstruct the original appearance of the image.
Digital watermarks are also capable of providing strong data protection (privacy) on line. Specific cryptographic methods prevent data access by unauthorized persons.
- Applicability in CCTV monitoring for automatically hiding sensitive parts of images (e.g. faces).
- Restoration of hidden parts of images by authorized person(s).
Identification of malicious image tampering
Digital images play an increasingly important role in our society. The ease of their acquisition and sharing makes them ubiquitous in many areas of life, ranging from holiday photos up to press photography and digital evidence. Prospective tampering with their content, which is becoming easier and more available, could have serious social, political or legal consequences. Hence, the ability to authenticate the content of digital images is a necessity in certain applications.
The INDECT project is working towards delivering new solutions and algorithms for efficient verification of the images' integrity. Our research focuses on the self-recovery capability, which allows to identify individual image fragments affected by malicious tampering, and to reconstruct their approximate original content. These features are enabled by a carefully designed digital watermark, which contains digital signatures of the content and roughly describes how the image should look like.
Our research focuses on the analysis of the behavior of such image authentication systems, their expected performance, and operation bounds. More information on the adopted approach can be found in:
 Paweł Korus and Andrzej Dziech, Efficient Method for Content Reconstruction with Self-Embedding, IEEE Transactions on Image Processing, (Volume:22, Issue: 3), 2013
Systems and procedures for information protection
As part of this work, standards and policies are being developed for secure access to information. The proposed procedures include secure authentication and authorization in accordance with generally accepted standards (especially X.509 certificates). In order to protect information against unauthorized access, cryptographic algorithms are being developed, and a new block cipher has been designed and implemented.
An innovative protocol for quantum cryptographic key distribution is also under development. Simulators have been created exploring the properties of encryption algorithms.
- Application of new cryptographic algorithms for enhancing information privacy and security.
- High level control and management of security mechanisms in quantum cryptography systems.
Intelligent search engine for court records
Intelligent search engine for court records using OCR technology and digital watermarks, solves technical problems in searching the content of court records. The search engine uses newly-designed dictionaries, reliable access control, and a high level of data security using digital watermarks. In the future, intelligent search may also be applied to judicial records in national registers and land registers.
For more information on INDECT tools and methods, please visit “The INDECT project – improving accuracy of existing video surveillance” – a European Commission webpage.