As AI continues to advance, we in the security industry must develop robust detection methods to preserve the accuracy and integrity of data, as this data is required to identify and respond to threats in the physical world. This is especially true on our continually advancing reliance on technology like video surveillance systems, VoIP intercom and phone systems, and protective intelligence used in security operations centers.

Understanding how to identify AI-generated content is crucial for maintaining trust and accuracy in an ever-evolving digital landscape. By staying vigilant and adapting detection strategies to keep pace with AI advancements, individuals and organizations can contribute to the ongoing battle against misinformation and deceptive practices, fostering a digital landscape built on trust and authenticity.

While the security industry progresses to embrace this technology in various capacities, let’s delve into some strategies that we mere mortals can use to detect AI’s presence in different domains, such as images, audio, and text.

While the security industry embraces AI in various capacities, understanding how to identify AI-generated content is crucial for maintaining trust and accuracy.

How to Detect AI in Images

Noise Analysis: AI-generated images often carry subtle patterns or artifacts that deviate from those present in naturally captured visuals. By scrutinizing noise patterns, and hard-to-generate items like hands, feet, logos, and background text in images, you can unveil the telltale signs of images crafted by AI algorithms. This nuanced examination aids in distinguishing between authentic and machine-generated visuals.

Examine the metadata: AI-generated images may lack the typical metadata markers associated with photos taken by traditional cameras. Anomalies in metadata such as geolocation, camera model, lens settings, etc., can serve as a red flag, prompting further investigation into the image’s authenticity.

Reverse image search: Employing reverse image search tools is a valuable technique to ascertain the origin of an image. AI-generated images often lack real-world counterparts, making them stand out in search results. This method is particularly effective in identifying instances where AI has been employed in the creation or manipulation of visual content.

How to Detect AI in Audio

Spectral analysis: AI-generated audio frequently exhibits specific spectral patterns or anomalies not found in naturally occurring sounds. Spectrogram analysis can unveil these hidden patterns, assisting in the identification of AI-generated audio content. This analytical approach is vital for differentiating between human and machine-generated sounds.

If you don’t have a spectrum analyzer handy, try these two human-capable detection methods. First, listen for words or phrases that get repeated – AI-generated content will use the same cadence, rhythm, and pronunciation, giving you a clue to its origin. Second, if you are having a dialog with AI, it does not know how to respond to bizarre questions – so if you get a call and it sounds like your family member asking for help and money, ask the AI for a code word like â€śwhat am I holding in between the toes of my left foot?” While the AI will have no idea how to respond, a family member will give you the secret code word or at least ask what the heck you are talking about.

Voiceprint analysis: Analyzing the voiceprint of a speaker can be instrumental in identifying AI-generated voices. AI often struggles to replicate the nuanced characteristics present in human speech, and voiceprint analysis can reveal discrepancies that may not be perceptible to the human ear.

Linguistic analysis: AI-generated speech may lack the natural flow and coherence found in human communication. Identifying unnatural syntax, grammar, or contextual inconsistencies can provide valuable clues in determining the authenticity of audio content.

How to Detect AI in Text

Linguistic anomalies: AI models may struggle with nuanced language use, so scrutinize for unnatural word choices, awkward sentence structures, or inconsistencies. Thorough linguistic analysis is crucial for revealing subtle signs of AI involvement in machine-generated text.

Stylometric analysis: AI models may exhibit consistent stylistic features that differ from the diversity found in human-generated content. Stylometric analysis, which focuses on patterns in writing style, is a powerful tool for detecting these subtle variations.

Contextual inconsistency: AI-generated text may struggle to maintain context throughout a piece of writing. Abrupt changes in tone, topic, or inconsistency in narrative flow can reveal instances where AI has been employed in the creation of textual content.

Cross-Domain Detection

Multi-modal Analysis: Combining techniques from image, audio, and text analysis creates a comprehensive approach to detecting AI across various domains. Cross-referencing results from different modalities can enhance the accuracy of AI detection and provide a more robust defense against machine-generated content.

Machine learning classifiers: Training machine learning classifiers using labeled datasets is an effective approach to distinguishing between AI-generated and human-created content. This method leverages the power of algorithms to identify subtle patterns across images, audio, and text, contributing to a more sophisticated and adaptive detection system.

Article Link –

This article originally appeared in the Ěý´Ç´ÚĚý. Paul F Benne is the President of ĐÓ°ÉÔ­´´ and has over 35 years in the protective service industry.

The post Tech Trends: How to Detect AI appeared first on ĐÓ°ÉÔ­´´.

The advancement of technology over the past three decades is astounding. Just a short 20 years ago, the vast majority of video surveillance systems were still using VHS tapes, multiplexers, and analog cameras with resolutions that were just a fraction of todays most basic of imaging technologies. Patrol, response, and investigations were exclusively done by manpower.

Today, we are seeing a massive influx of true Artificial Intelligence – albeit mostly basic video-based analytics – into these traditional security technologies.

Basic AI in Practice for Security

We are now able to see the emergence of AI’s most elementary abilities surfacing in the security industry. An example is Coresight.ai, whose technology is able to provide facial identification of individuals using standard video surveillance cameras – even when the face of the subject is partially obscured. The technology provides the ability to match a person of interest on a watch list to individuals captured on live and recorded video.

Consider Zeroeyes and Omnilert, whose AI technologies use standard video surveillance cameras to detect a person in possession of a gun, prompting the system to trigger an action or alert prior to the subject carrying the gun making entry to a facility or shooting the weapon.

Companies like Viisights bring behavioral analytics that are able to detect a wide variety of behaviors like fighting, crowd formation, accidents, slip-and-falls, throwing objects, or other aggressive behaviors.

The vast amount of currently installed security technology is wholly reactive, meaning that an incident must take place prior to a mitigative action being taken. These technologies make security systems and departments proactive in a way we have never seen. AI can greatly speed the detection and response to many incident types, and also trigger on precursors that happen before an actual incident occurs.

As for AI and robotics, by now you have likely seen Spot the robot dog from Boston Dynamics, and perhaps some of you have seen Atlas, the most dynamic humanoid robot in the world from the same company. There have been plenty of similar robots roaming tradeshow floors the past couple of years.These robots now have human-like agility and the ability to analyze their environment. They are equipped with an array of sensors that give the robotic platform detailed information on its movement, surroundings, and position, as well as process incredible amounts of data in milliseconds. The robot can calculate thousands of probable scenarios, weighing what ones have the most likelihood of occurring, then almost instantly react to the situation, faster than humans are capable.

Imagine the Possibilities

You can easily extrapolate the possibilities. Consider the integration and convergence of just these few AI technologies and robotics mentioned: A person with a gun exits their vehicle at a school or bank, the AI determines the presence of a gun and also determines that the person in possession of the gun is on the facility’s watch list (placed on the list perhaps because they are a terminated employee or a student identified as high risk), this triggers an automated lockdown, pre-scripted announcements, law enforcement notifications, starts live video streaming to responding agencies, and automatically tracks the location and actions of the subject. Then a security robot, powered with this real-time information, moves into place to intercept the subject as they attempt to gain entry into the facility.

The robot can analyze threats, verbal exchanges, micro expressions and movements of the subject against thousands of like scenarios or previously documented incidents to make a determination of the subject’s probable actions before the subject even takes them. Then with faster-than-human reaction time and agility, engage the subject with less-than-lethal force to disarm and subdue the subject.

While we are not quite there yet, it is certainly on the horizon. In April, the New York Police Department announced the use of Spot the robot dog and other robots to start patrolling Times Square.

Playing Devil’s Advocate

To say this article is far-reaching would be a false statement; so why would I choose to open with such a negative statement like “the total destruction of mankind?”

Let’s entertain another scenario, an adversarial one. We know that all technology eventually becomes more commonly understood and can be used for both good and bad. Currently, the development of AI and robotics technology has little to no regulation or oversight, providing few limits to what this technology can be programmed to do, or better yet, to not do.

Elon Musk is quoted as saying AI is “quite a dangerous technology” and that AI has the potential to destroy civilization. What happens when an adversary – criminal enterprise, enemy state, or the AI itself – obtains and programs AI to do their bidding?

Imagine an adversary who develops an integrated AI robot with advanced sensory technology and physical agility, the ability to analyze and process petaflops of information to render an action hundreds of times faster than any human’s ability. The outcome is potentially unstoppable. It is a frightful view into a “Terminator” movie-type threat that security and law enforcement teams are not remotely ready or equipped to respond to.

Avoiding the Ugly

We should not be surprised that the security industry is embracing the newest emerging technologies such as AI and robotics. As illustrated, they are an obvious force multiplier and bring new efficiencies to improve the security function.

However, the bad is there, just bubbling under the surface.

In the end, the security industry should embrace a better approach to develop and use this technology with some firm ground rules – to hopefully harness the good and avoid the bad and potentially ugly.

Article Link –

This article originally appeared in the Ěý´Ç´ÚĚý. Paul F Benne is the President of ĐÓ°ÉÔ­´´ and has over 35 years in the protective service industry.

The post Tech Trends: AI – The Good, the Bad, and Perhaps the Ugly appeared first on ĐÓ°ÉÔ­´´.

If you are considering this as a provider or an end-user, you will need to look beyond the sales pitch to see the true value – and the drawbacks – of this technology trend.

At first pass, the “aaS” model can be very annoying to me as a consumer. I used to go into a store or browse online at the software available and buy it outright. If I elected to upgrade to a newer version in the future I could, but my hand was seldom forced, and the power was all mine.

Now, you are hard-pressed to find a software platform that is not offered in some type of subscription or SaaS model, and this trend is moving from the software world to hardware, systems and services. It is like you cannot “buy” anymore, only rent, but is this all bad?

The Pros and Cons
Understanding some of the pros and cons of this trend can help you understand its value, both to yourself individually and to an overall organization. While there are many aspects of the “aaS” sales model, here are five considerations you should evaluate prior to making the commitment – keeping in mind that one person’s pro may be another person’s con. Thus, consider these points under your own application and circumstances.

1. Limits expensive server hardware: When using platforms like Ava Security, Brivo, OpenEye, IronYun, etc. certain configurations of these services can be cloud-based. This can reduce or eliminate the need for on-premises servers, which are costly and have a limited lifespan.

If your customer is considering a traditional deployment of security technology, they need to consider that the servers installed on premises will come at a high cost and will need to be replaced in 5-8 years. They will also need to consider that they will likely need a Software Service Agreement (SSA) and a preventive maintenance contract to keep the system updated with patches, fixes and version releases. With an “aaS” model, all of this is included.

2. CapEx vs OpEx: While evaluating “aaS” options, take into consideration how the customer will buy the technology. Many construction projects have deeper CapEx budgets to allow for the initial procurement and installation of expensive security technology; however, buying “aaS” typically gets expensed through OpEx budgets and is a recurring monthly or annual expense – which can be more highly scrutinized. Having a thorough understanding of how “aaS” is funded and working out the details with the end-user’s finance department is critical to resolving push-back regardless of what solution is up for consideration.
3. Critical facilities, SOCs and uptime requirements: For facilities that operate in a critical capacity and have an on-site SOC, GSOC, or other real-time point of aggregating security information, close attention to “aaS” features are mandatory. Most “aaS” offerings have some or all features that are entirely dependent on outbound network connectivity; therefore, if the network fails, the security operation loses the ability to access and use the “aaS” product.

When evaluating “aaS,” ask your customer very pointed and hard questions about what services operate and if there are critical services that will not work in the event that outbound network connectivity is severed.

4. Bandwidth and network: The “aaS” sales pitch is typically a good one when it is well-perfected and delivered, but there are some downsides to consider, and integrators will need the end-user’s IT department or a non-biased third party like a consultant to help navigate the sales pitch and dig into the tech requirements of the network.

Having 10, 50, 100, or 1000 cameras at a facility that are going to a cloud-based “aaS” will have big implications on network architecture and bandwidth. Be sure this is fully understood before making a commitment to the “aaS” technology.

5. Proprietary equipment: Some “aaS” platforms openly accept cameras and edge technology from multiple manufacturers; others have cameras that can be used on other systems should you elect to stop using those systems in the future. That said, be warned that this is not the case for all “aaS” providers; in fact, some providers’ cameras are as good as a brick should your customer stop using the “aaS” platform. This could be a costly and embarrassing mistake.

Best Practices
Here are key takeaways for considering any type of “aaS” model for software, systems, or services:

Thoroughly understand what your customer’s needs are before calling “aaS” companies in to do a sales pitch. Not having this will distract you with sales presentation sparkles, bells, and whistles that may be great features, but can deviate from the core performance objectives required to serve the business.

Understand the points of failure in “aaS” models and be willing to accept the risks they present. Knowing exactly where the points of failure are enables the integrator and end-user to build operational processes to stopgap the vulnerability should a failure occur.

Know the total cost of ownership (TCO) of both the “aaS” and traditional deployment of the software, systems, and services you are evaluating.

Don’t go it alone. Bring in IT, Finance, and other key stakeholders of the business to help evaluate the solution from their areas of expertise.

Paul F. Benne is a 35-year veteran in the protective services industry. He is President of ĐÓ°ÉÔ­´´ LLC, a security consulting and design firm in based in New York City. Connect with him at www.linkedin.com/in/paulbenne or visit www.sentinelconsulting.us.

Article Link – https://www.securityinfowatch.com/integrators/article/21278024/tech-trends-security-tech-for-nonsecurity-stakeholders

This article originally appeared in the November 2022 issue of Security Business magazine. Paul F Benne is the President of ĐÓ°ÉÔ­´´ and has over 35 years in the protective service industry.

The post Tech Trends: Everything as a Service appeared first on ĐÓ°ÉÔ­´´.

Communication: The members of the team must be willing to communicate ideas and issues with one another.  Conversations must take place in which everyone listens to the speaker and feels free to respond to what was said. Team members should approach all conversations with respect, trust and honesty. They should all feel like they have a voice and their opinion matters. There is no I in T.E.A.M.

Sense of Responsibility:  Each member of the protection team should have a sense of responsibility to the team and to the location they are protecting. A protection team can be successful when team members understand and live up to their responsibilities to the team.

Know Your Role: A successful team must have clearly defined roles for each member. Michael Jordan is arguably the best basketball player of all time but he couldn’t score every point, grab every rebound or defend every position. He needed help from his teammates and those members had to know their role. Defining roles helps team members to understand their responsibility to the team and will allow them to understand the roles and responsibilities of their team members as well.

Ability to Resolve Conflicts; All teams experience some kind of conflict. A successful team will be able to work through issues that arise within the team. There is a saying that states; “Teeth and tongue fall out sometime, which means that people who are closest to each other can have disagreements, but they still have to work together for a common purpose. A protection team will spend a lot of time together during the course of a workday. Allowing conflicts to linger within the team will have a negative effect on the overall goal which is to provide protection for people, assets and property.

Support from leadership: Like a coach setting the pace for the team, a person in leadership needs to recognize and uphold the responsibility to lead their team with support in both the team’s success and failures. This lets a team know that they are in it for the win. Knowing that leadership stands behind them and supports them when they need it is a tool of empowerment.

When a team works together goals can be achieved. There are no big I’s or little U’s because Together Everyone Achieves More.

The post T.E.A.M.: Together Everyone Achieves More appeared first on ĐÓ°ÉÔ­´´.

ĐÓ°ÉÔ­´´’s approach to security design is a holistic one.  It focuses on the what and how of risk mitigation.  It is beyond a best practices approach, in that countermeasures are not selected and deployed simply because its a standard or general industry practice.  Instead, a risk-based approach is used.  All aspects of a Client’s operation are considered and the risks are studied.  Countermeasures are then selected and deployed to mitigate these observed risks.

This is where the relationship between our Security Consultants and Security Engineers becomes so apparent.  Our Security Consultants specialize in identifying threats and vulnerabilities against given assets, resulting in a risk profile.  This is the WHAT and WHY.

Security Engineers then review this risk profile with our Security Consultants and determine what countermeasures are appropriate, whether they be physical or electronic, or a combination thereof.  This is the HOW.

WHAT and WHY are we protecting? HOW are we going to protect it?

These questions form the basis of every design that ĐÓ°ÉÔ­´´ produces.  Security Engineers cannot alone determine the proper countermeasures, where there needed, and why they’re needed. Security Consultants cannot alone determine how physical and electronic security countermeasures can be deployed and managed.  Only working together, and combining knowledge and experience, can a successful security program be implemented.

The Security Consultants at ĐÓ°ÉÔ­´´ have in depth, real world knowledge of security operations, emergency response, and understand how countermeasures are used to mitigate risks.

The Security Engineers at ĐÓ°ÉÔ­´´ are educated engineers with knowledge and experience in a variety of physical and electronic countermeasures, their design requirements, deployment, and how they are operated.

Don’t simply lock your front door and back door because its a best practice.  Consider your risks, and mitigate those risks properly.  Simply placing a lock on a door may not be adequate protection for your most valued assets.  What assets do you have that require specific mitigations?

The post The Relationship between Security Consultants and Security Engineers appeared first on ĐÓ°ÉÔ­´´.