Data storage devices across the security industry are routinely required to handle an enormous amount and many layers of raw data. As Safe City projects in varying sizes become more prevalent, the number of surveillance nodes has reached the hundreds of thousands. And due to the widespread use of high-definition monitoring, the amount of data involved in security surveillance has increased dramatically in a short time. Efficient collection, analysis, and application of data and the intelligent use of it are becoming ever more critical in this industry. Thus, improving video intelligence appears to be an inevitable, industry-wide goal.

Security users hope that their investment in new products will bring even more benefits beyond simply tracing and tracking persons of interest and evidence collection after a security event. Some examples of added benefits include using the latest technologies to replace the large amount of man-power previously required for searching surveillance footage, detecting anomalous data, and finding ever more efficient ways to allow surveillance to shift from post-incident tracing to alerts during incidents—or even pre-incident alerts. In order to satisfy these demands, new technologies are required. Intelligent video surveillance has been available for many years. However, the outcomes of its application have not been ideal. The emergence of deep learning has enabled these demands to become reality.

The insufficiency of traditional intelligent algorithms

Traditional intelligent video surveillance has especially strict requirements for a scene’s background. The accuracy of intelligent recognition and analysis in comparable scenarios remains inconsistent. This is primarily due to the fact that traditional intelligent video analysis algorithms still have many flaws. In an intelligent recognition and analysis process, such as human facial recognition, two key steps are required: First, features are extracted, and second, “classification learning” is performed.

The features in traditional intelligent algorithms are designed by humans and have always been heavily subjective

The degree of accuracy in this first step directly determines the accuracy of the algorithm. In fact, most of the system’s calculation and testing workload is consumed in this part. The features in traditional intelligent algorithms are designed by humans and have always been heavily subjective. More abstract features—those that humans have difficulty comprehending or describing—are inevitably missed. With shifting angles and lighting, and especially when the sample size is enormous, many features can be too difficult to detect. Therefore, while traditional intelligent algorithms perform well in very specific environments, subtle changes (image quality, environment, etc.) yield significant challenges to accuracy.

Target detection and attribute recognition

The second step—classification learning—mainly involves target detection and attribute recognition. As the number of available categories for classification rises, so does the difficulty level. Hence, traditional intelligent analysis technologies are highly accurate in vehicle analysis but not in human and object analysis. For example, in vehicle detection, a distinction is made between a vehicle and a non-vehicle, so the classification is simple and the level of difficulty is low. To recognise vehicle attributes requires recognition of different vehicle designs, logos, and so on. However, there are relatively few of these, making the classification results generally accurate. On the other hand, if recognition is to be performed on human faces, each person is a classification of its own, and the corresponding categories will be extremely numerous—naturally leading to a very high level of difficulty.

The accuracy of intelligent recognition and analysis in comparable scenarios remains inconsistent
Enhanced accuracy is the result of multi-layer learning and extensive data collection

Traditional intelligent algorithms generally use shallow learning models to handle situations with large amounts of data in complex classifications. The analysis results are far from ideal. Furthermore, these results directly restrict the breadth and depth of intelligent applications and further development. Hence the need for increasing the “depth” of intelligence in big data for the security industry is arising.

The advantages of Deep Learning and its algorithms

Traditional intelligent algorithms are designed by humans. Whether or not they are designed well depends greatly on experience and even luck, and this process requires a lot of time. So, is it even possible to get machines to automatically learn some of the features? Yes! This is actually the objective of Artificial Intelligence (AI).

The inspiration for deep learning comes from a human brain’s neural networks. Our brains can be seen as a very complex deep learning model. Brain neural networks are comprised of billions of interconnected neurons; deep learning simulates this structure. These multi-layer networks can collect information and perform corresponding actions. They also possess the ability for object abstraction and recreation. Deep learning is intrinsically different from other algorithms. The way it solves the insufficiencies of traditional algorithms is encompassed in the following aspects.

Algorithmic model for deep learning

The algorithmic model for deep learning has a much deeper structure than the two 3-layered structures of traditional algorithms. Sometimes, the number of layers can reach over a hundred, enabling it to process large amounts of data in complex classifications. Deep learning is very similar to the human learning process, and has a layer-by-layer feature-abstraction process. Each layer will have different “weighting,” and this weighting reflects on what was learned about the images’ “components.” The higher the layer level, the more specific the components. Simulating the human brain, an original signal in deep learning passes through layers of processing; next, it takes a partial understanding (shallow) to an overall abstraction (deep) where we can perceive the object.

Deep learning does not require manual intervention but relies on a computer to extract features by itself. This way it is able to extract as many features from the target as possible, including abstract features that are difficult or impossible to describe. The more features there are, the more accurate the recognition and classification will be. Some of the most direct benefits that deep learning algorithms can bring include achieving comparable or even better-than-human pattern recognition accuracy, strong anti-interference capabilities, and the ability to classify and recognise thousands of features.

Hikvision has operated in the security industry for many years with its own research and development capabilities

Key factors of Deep Learning

In total, there are three main reasons why deep learning only became popular in recent years and not earlier: the scale of data involved, computing power, and network architecture.

Improvements in data-driven algorithm performance have accelerated deep learning in various intelligent applications in a short amount of time. Specifically, with the increase in data scale, algorithmic performance improved as well. Accordingly, user experience has improved and more users are involved, further facilitating a larger scale of data.

Video surveillance data makes up 60% of big data, and the amount is rising at 20% annually. The speed and scale of this achievement is due to the popularisation of high definition video surveillance—HD 1080p is becoming more common, and 4K and higher resolutions are gradually being applied in many important applications.

Hikvision has operated in the security industry for many years with its own research and development capabilities, employing large amounts of real video and image data as training samples. With a large amount of good quality data, and over a hundred team members to label the video images, sample data with millions of categories have been accumulated. With this large amount of quality training data, human, vehicle, and object pattern recognition models will become more and more accurate for video surveillance use.

The deep learning model requires a large amount of samples, making a large amount of calculations inevitable
Enhanced accuracy is the result of multi-layer learning and extensive data collection

Higher computational power

Furthermore, high performance hardware platforms enable higher computational power. The deep learning model requires a large amount of samples, making a large amount of calculations inevitable. In the past, hardware devices were incapable of processing complex deep learning models with over a hundred layers. In 2011, Google’s DeepMind used 1,000 devices with 16,000 CPUs to simulate a neural network with approximately 1 billion neurons. Today, only a few GPUs are required to achieve the same sort of computational power with even faster iteration. The rapid development of GPUs, supercomputers, cloud computing, and other high performance hardware platforms has allowed deep learning to become possible.

Finally, the network architecture plays its own role in advancing deep learning. Through constant optimisation of deep learning algorithms, better target-object recognition can be achieved. For more complex applications such as facial recognition or in scenarios with different lighting, angles, postures, expressions, accessories, resolutions, etc., network architecture will impact the accuracy of recognition, i.e., the more layers in deep learning algorithms, the better the performance.

In 2016, Hikvision achieved the number one position in the Scene Classification category at the ImageNet Large Scale Visual Recognition Challenge 2016. The team from Hikvision Research Institute used inception-style networks and not-so-deep residual networks that perform better in considerably less training time, according to Hikvision’s experiments for training and testing.

Furthermore, Hikvision’s Optical Character Recognition (OCR) Technology, based on Deep Learning and led by the company’s Research Institute, also won the first price in the ICDAR 2016 Robust Reading Competition. The Hikvision team substantially surpassed both strong domestic and foreign competitors in three word-recognition challenges, including born-digital images, focused scene text, and incidental scene text, demonstrating that the word recognition technology by Hikvision reached the world’s top level.

In the past two years, deep learning technology has excelled in speech recognition, computer
vision, voice translation,
and much more

Application of Deep Learning products

In the past two years, deep learning technology has excelled in speech recognition, computer vision, voice translation, and much more. It has even surpassed human capabilities in the areas of facial verification and image classification; hence, it has been highly regarded in the field of video surveillance for the security industry.

In the application of intelligent video in target detection, tracking, and recognition, the rise of deep learning has had a profound influence. When applying those three functions, deep learning potentially touches upon every aspect of the security video surveillance industry: facial detection, vehicle detection, non-motor vehicle detection, facial recognition, vehicle brand recognition, pedestrian detection, human body feature detection, abnormal facial detection, crowd behaviour analysis, multiple target tracking, and so on.

These types of intelligent functions require a series of front-end surveillance cameras, back-end servers and other products which support deep learning algorithms. In small scale applications, front-end cameras can directly operate structured human and vehicle feature extraction, and tens of thousands of human facial images can be stored within the front-end devices to implement direct facial comparison, so as to reduce costs of communicating with a server. In large scale applications, front-end cameras can work with back-end servers. Specifically, the structured video task is handled by front-end devices, reducing the workload for back-end devices; matching and searching efficiency of back-end servers improve as well.

Hikvision new products with Deep Learning

This year, Hikvision will soon introduce a series of products with deep learning technology, such as the DeepInview Series cameras which can accurately detect, recognise, and analyse human, vehicle, and object features and behaviour, and can be widely used in indoor and outdoor scenarios. Another of products worth mentioning is Hikvision’s DeepInmind Series of NVRs which incorporate advanced deep learning algorithms and imitate human thoughts and memory. The DeepInmind products feature an innovative NVR+GPU mode, retaining the advantages of traditional NVRs and additional structured video analysis functions, which together greatly improve the value of video.

Deep learning is the next level of AI development. It is beyond machine learning where supervised classification of features and patterns are set into algorithms. Deep learning incorporates unsupervised or “self-learning” principles. Hikvision is developing this concept in its own analytics algorithms. Enhanced accuracy is the result of multi-layer learning and extensive data collection. Application of this algorithm into face recognition, vehicle recognition, human recognition, and other platforms will significantly advance the performance of analytics.

Download PDF version

In case you missed it

Adapting servers for IP video surveillance systems: Why manufacturers struggle
Adapting servers for IP video surveillance systems: Why manufacturers struggle

Security integrators are often tasked with a multitude of responsibilities which could include a variety of installation, integration or design tasks made up of sprinkler systems, fire alarms, access control, HVAC, video surveillance systems and networks; and then pile on maintenance, training and analytics. Traditionally, most security integrators have installation backgrounds but are now expected to be IT savvy, too. Even the most proficient IT professionals may not fully grasp the complexity of adapting computer servers for use with video systems. It’s not the area of expertise of security integrators as the complexities between IT data and video data are significant. Therefore, security integrators depend on system builders to provide solutions to meet the needs of video systems expertly and with few hassles. It’s a simple enough ask, but not so easy to deliver. Tom Larson, Chief Technology Officer, BCDVideo, lists some of the challenges: The gap between reality and customer expectations End users should expect a security integrator to provide services and a wide product line to ensure the right equipment for any size job Sometimes there is a gap between what a security integrator expects from a video surveillance solution (in terms of validation testing, dependability, technical support) and the performance of available choices, especially in the case of low-cost or generic equipment. Extra service and support are needed to bridge the gap. Unfortunately, some manufacturers entering the market have failed to deliver, and integrators (and their end user customers) have paid a price. The network is often overlooked Security integrators should pay special attention to engineering the network and calculating the bandwidth and storage needed for video projects, especially given how technology evolves so quickly. Security is an appliance-driven business, and integrators who just want to add another server to expand storage or functionality without configuring the network run the risk of i/o bottlenecks and other system failures. End users should expect a security integrator to provide services and a wide product line to ensure the right equipment for any size job. Unfortunately, traditional IT resellers are often married to a singular solution limiting their knowledge of a good fit for the job. Buying a video server based on a low price aggravates the problem, as “Frankensteined” or generic servers tend to generate additional costs over time Servers are mistakenly considered a one-time expense One mistake purchasing agents make and security integrators have a hard time quantifying is viewing video storage as a capital expense (as one more component of a security system) rather than considering ongoing operating expenses. Buying a video server based on a low price aggravates the problem. “Frankensteined” or generic servers tend to generate additional costs over time, such as firmware or supply chain issues, and some systems builders have failed to provide support to offset those costs. In fact, the high costs over time of supporting inexpensive servers have been unsustainable for some system builders, who have left integrators and end users holding the bag, and in some cases, the liability. Adapting to sustainable strategies “Systems builders to the video surveillance market must adapt and invest to meet the demands of security integrators’ expectations, and they need a business model that enables them to provide a substantial level of support and commitment,” says Larson. “Working with high-quality manufacturers and providing tried-and-tested, certified equipment upfront ensures manageable costs over the life of the system. Products that are fully tested and contain no firmware bugs ensure smoother installations. By providing adequate technical support to the security integrator and managing IT variables over the life of the system, the systems builder makes it possible for a security integrator to specify and install a video server as easily as any other system component.” Keeping IT professionals on staff to deal with server issues is cost-prohibitive for security integrators Taking a longer-term view and considering total cost of ownership is a more sustainable strategy for integrators, says Larson. Investing upfront in a higher-quality server is rewarded by dependability and lower service costs over the life of the system. And the lower costs of supporting a higher-quality server create a more sustainable business model for the integrator, thus ensuring the integrator and end user will have ongoing support. Adapting server technology to video applications Security integrators deliver a different skill set than IT integrators, who tend to be more hands-on in terms of updating firmware and providing maintenance. Keeping IT professionals on staff to deal with server issues is cost-prohibitive for security integrators, who therefore depend on systems builders to provide that expertise. They develop a long-term relationship with a systems builder they can depend on to meet their needs for each job. Larson says the best scenario for a security integrator is a combination of a high-quality server systems builder that understands the specific needs of the security integrator market. Adapting server technology to video applications requires knowledge of both disciplines. Dependable technology adapted to the needs of the video channel ensures successful installations and happy, long-term customers.

New Year’s Resolutions to counter web and mobile application security breaches in 2019
New Year’s Resolutions to counter web and mobile application security breaches in 2019

With the coming of a New Year, we know these things to be certain: death, taxes, and… security breaches. No doubt, some of you are making personal resolutions to improve your physical and financial health. But what about your organisation’s web and mobile application security? Any set of New Year’s resolutions is incomplete without plans for protecting some of the most important customer touch points you have — web and mobile apps. Every year, data breaches grow in scope and impact. Security professionals have largely accepted the inevitability of a breach and are shifting their defense-in-depth strategy by including a goal to reduce their time-to-detect and time-to-respond to an attack. Despite these efforts, we haven’t seen the end of headline-grabbing data breaches like recent ones affecting brands such as Marriott, Air Canada, British Airways and Ticketmaster. App-level threats The apps that control or drive these new innovations have become today’s endpoint The truth of the matter is that the complexity of an organisation’s IT environment is dynamic and growing. As new technologies and products go from production into the real world, there will invariably be some areas that are less protected than others. The apps that control or drive these new innovations have become today’s endpoint — they are the first customer touch point for many organisations. Bad actors have realised that apps contain a treasure trove of information, and because they are often left unprotected, offer attackers easier access to data directly from the app or via attacks directed at back office systems. That’s why it’s imperative that security organisations protect their apps and ensure they are capable of detecting and responding to app-level threats as quickly as they arise. It’s imperative that security organisations protect their apps and ensure they are capable of detecting and responding to app-level threats as quickly as they arise In-progress attack detection Unfortunately, the capability to detect in-progress attacks at the app level is an area that IT and security teams have yet to address. This became painfully obvious in light of the recent Magecart attacks leveraged against British Airways and Ticketmaster, among others. Thanks to research by RiskIQ and Volexity, we know that the Magecart attacks target the web app client-side. During a Magecart attack, the transaction processes are otherwise undisturbed Attackers gained write access to app code, either by compromising or using stolen credentials, and then inserted a digital card skimmer into the web app. When customers visited the infected web sites and completed a payment form, the digital card skimmer was activated where it intercepted payment card data and transmitted it to the attacker(s). Data exfiltration detection During a Magecart attack, the transaction processes are otherwise undisturbed. The target companies receive payment, and customers receive the services or goods they purchased. As a result, no one is wise to a breach — until some 380,000 customers are impacted, as in the case of the attack against British Airways. The target companies’ web application firewalls and data loss prevention systems didn’t detect the data exfiltration because those controls don’t monitor or protect front-end code. Instead, they watch traffic going to and from servers. In the case of the Magecart attacks, the organisation was compromised and data was stolen before it even got to the network or servers. Today’s proven obfuscation techniques can help prevent application reverse engineering, deter tampering, and protect personal identifiable information and API communications Best practice resolutions The Magecart attacks highlight the need to apply the same vigilance and best practices to web and mobile application source code that organisations apply to their networks—which brings us to this year’s New Year’s resolutions for protecting your app source code in 2019: Alert The key to success is quickly understanding when and how an app is being attacked First, organisations must obtain real-time visibility into their application threat landscape given they are operating in a zero-trust environment. Similar to how your organisation monitors the network and the systems connected to it, you must be able to monitor your apps. This will allow you to see what users are doing with your code so that you can customise protection to counter attacks your app faces. Throughout the app’s lifecycle, you can respond to malicious behavior early, quarantine suspicious accounts, and make continuous code modifications to stay a step ahead of new attacks. Protect Next, informed by threat analytics, adapt your application source code protection. Deter attackers from analysing or reverse engineering application code through obfuscation. Today’s proven obfuscation techniques can help prevent application reverse engineering, deter tampering, and protect personal identifiable information and API communications. If an attacker tries to understand app operation though the use of a debugger or in the unlikely event an attacker manages to get past obfuscation, threat analytics will alert you to the malicious activity while your app begins to self-repair attacked source code or disable portions of the affected web app. The key to success is quickly understanding when and how an app is being attacked and taking rapid action to limit the risk of data theft and exfiltration. Protecting encryption keys is often overlooked but should be considered a best practice as you forge into the new year with a renewed commitment to app security to ensure your organisation’s health and well-being in 2019 Encrypt Finally, access to local digital content and data, as well as communications with back office systems, should be protected by encryption as a second line of defense, after implementing app protection to guard against piracy and theft. However, the single point of failure remains the instance at which the decryption key is used. Effective encryption requires a sophisticated implementation of White-Box Cryptography This point is easily identifiable through signature patterns and cryptographic routines. Once found, an attacker can easily navigate to where the keys are constructed in memory and exploit them. Effective encryption requires a sophisticated implementation of White-Box Cryptography. One that combines a mathematical algorithm with data and code obfuscation techniques transforming cryptographic keys and related operations into indecipherable text strings. Protecting encryption keys is often overlooked but should be considered a best practice as you forge into the new year with a renewed commitment to app security to ensure your organisation’s health and well-being in 2019. Protecting applications against data breach According to the most recent Cost of a Data Breach Study by the Ponemon Institute, a single breach costs an average of $3.86 million, not to mention the disruption to productivity across the organisation. In 2019, we can count on seeing more breaches and ever-escalating costs. It seems that setting—and fulfilling—New Year’s resolutions to protect your applications has the potential to impact more than just your risk of a data breach. It can protect your company’s financial and corporate health as well. So, what are you waiting for?

How organisations can secure user credentials from data breaches and password hacks
How organisations can secure user credentials from data breaches and password hacks

In the age of massive data breaches, phishing attacks and password hacks, user credentials are increasingly unsafe. So how can organisations secure accounts without making life more difficult for users? Marc Vanmaele, CEO of TrustBuilder, explains. User credentials give us a sense of security. Users select their password, it's personal and memorable to them, and it's likely that it includes special characters and numbers for added security. Sadly, this sense is most likely false. If it's anything like the 5.4 billion user IDs on haveibeenpwned.com, their login has already been compromised. If it's not listed, it could be soon. Recent estimates state that 8 million more credentials are compromised every day. Ensuring safe access Data breaches, ransomware and phishing campaigns are increasingly easy to pull off. Cyber criminals can easily find the tools they need on Google with little to no technical knowledge. Breached passwords are readily available to cyber criminals on the internet. Those that haven’t been breached can also be guessed, phished or cracked using one of the many “brute-force” tools available on the internet. It's becoming clear that login credentials are no longer enough to secure your users' accounts. Meanwhile, organisations have a responsibility and an ever-stricter legal obligation to protect their users’ sensitive data. This makes ensuring safe access to the services they need challenging, particularly when trying to provide a user experience that won’t cause frustration – or worse, lose your customers’ interest. After GDPR was implemented across the European Union, organisations could face a fine of up to €20 million, or 4% annual global turnover Importance of data protection So how can businesses ensure their users can safely and simply access the services they need while keeping intruders out, and why is it so important to strike that balance? After GDPR was implemented across the European Union, organisations could face a fine of up to €20 million, or 4% annual global turnover – whichever is higher, should they seriously fail to comply with their data protection obligations. This alone was enough to prompt many organisations to get serious about their user’s security. Still, not every business followed suit. Cloud security risks Breaches were most commonly identified in organisations using cloud computing or where staff use personal devices According to a recent survey conducted at Infosecurity Europe, more than a quarter of organisations did not feel ready to comply with GDPR in August 2018 – three months after the compliance deadline. Meanwhile, according to the UK Government’s 2018 Cyber Security Breaches survey, 45% of businesses reported breaches or attacks in the last 12 months. According to the report, logins are less secure when accessing services in the cloud where they aren't protected by enterprise firewalls and security systems. Moreover, breaches were most commonly identified in organisations using cloud computing or where staff use personal devices (known as BYOD). According to the survey, 61% of UK organisations use cloud-based services. The figure is higher in banking and finance (74%), IT and communications (81%) and education (75%). Additionally, 45% of businesses have BYOD. This indicates a precarious situation. The majority of businesses hold personal data on users electronically and may be placing users at risk if their IT environments are not adequately protected. Hackers have developed a wide range of tools to crack passwords, and these are readily available within a couple of clicks on a search engine Hacking methodology In a recent exposé on LifeHacker, Internet standards expert John Pozadzides revealed multiple methods hackers use to bypass even the most secure passwords. According to John’s revelations, 20% of passwords are simple enough to guess using easily accessible information. But that doesn’t leave the remaining 80% safe. Hackers have developed a wide range of tools to crack passwords, and these are readily available within a couple of clicks on a search engine. Brute force attacks are one of the easiest methods, but criminals also use increasingly sophisticated phishing campaigns to fool users into handing over their passwords. Users expect organisations to protect their passwords and keep intruders out of their accounts Once a threat actor has access to one password, they can easily gain access to multiple accounts. This is because, according to Mashable, 87% of users aged 18-30 and 81% of users aged 31+ reuse the same passwords across multiple accounts. It’s becoming clear that passwords are no longer enough to keep online accounts secure. Securing data with simplicity Users expect organisations to protect their passwords and keep intruders out of their accounts. As a result of a data breach, companies will of course suffer financial losses through fines and remediation costs. Beyond the immediate financial repercussions, however, the reputational damage can be seriously costly. A recent Gemalto study showed that 44% of consumers would leave their bank in the event of a security breach, and 38% would switch to a competitor offering a better service. Simplicity is equally important, however. For example, if it’s not delivered in ecommerce, one in three customers will abandon their purchase – as a recent report by Magnetic North revealed. If a login process is confusing, staff may be tempted to help themselves access the information they need by slipping out of secure habits. They may write their passwords down, share them with other members of staff, and may be more susceptible to social engineering attacks. So how do organisations strike the right balance? For many, Identity and Access Management solutions help to deliver secure access across the entire estate. It’s important though that these enable simplicity for the organisation, as well as users. Organisations need an IAM solution that will adapt to both of these factors, providing them with the ability to apply tough access policies when and where they are needed and prioritising swift access where it’s safe to do so Flexible IAM While IAM is highly recommended, organisations should seek solutions that offer the flexibility to define their own balance between a seamless end-user journey and the need for a high level of identity assurance. Organisations’ identity management requirements will change over time. So too will their IT environments. Organisations need an IAM solution that will adapt to both of these factors, providing them with the ability to apply tough access policies when and where they are needed and prioritising swift access where it’s safe to do so. Importantly, the best solutions will be those that enable this flexibility without spending significant time and resource each time adaptations need to be made. Those that do will provide the best return on investment for organisations looking to keep intruders at bay, while enabling users to log in safely and simply.