The rapid growth of Artificial Intelligence (AI) has completely revolutionized how the world works in today’s date. With several industries now shifting to AI, it is of the utmost importance to make the technology foolproof.
To understand why AI needs to be secured, we have to first understand what the term represents. In a nutshell, AI enables machines to function in an “intelligent” manner and use their own “brains”. It can undertake tasks that generally require human intervention. For instance, it can perform several functions such as decision making, visual perception, speech recognition, and language translation to name amongst a few.
To know more about AI, you can check out this blog - What is AI
The German-headquartered global brand Bosch has established a brand name for itself since its inception in 1889 as one of the world’s leading global suppliers of technology and services, IT and Business solutions, and end-to-end engineering. In an exclusive interview with Analytics Steps, Mr. Manojkumar Parmar, Deputy General Manager at Bosch sheds light on the development of AI and the security aspects of AI.
An IIM Bengaluru alumni, Parmar has worked proactively with the multinational brand for 13 years and has filed a series of patents in his journey.
AI – A human innovation
Every time a scientist invents an instrument, it becomes a gadget for the masses to use. Parmar explains it through an example, “For example, earlier Logarithms Tables were used to multiply a large number, it was then replaced by the handheld calculator which came into existence in 1967. The calculators were soon replaced by modern computers in the ’80s. With the turn of the century, smartphones & tablets took on the role of the computers and currently the focus is on AI.”
The term Artificial Intelligence was coined in 1956 by John McCarthy during a conference at Dartmouth College, New Hampshire. Though it would still take years for the idea of AI to formally come into existence, the blueprint for its inception was laid.
The history of AI is a long, interesting one. After decades of research by several countries, especially the United States and Japan, finally, AI came into the limelight. In 1997, IBM’s Deep Blue computer defeated Russian Grand Master Garry Kasparov, the same year Dragon Systems developed a speech recognition software that was used in Microsoft Windows. With better data storage capabilities and cheaper manufacturing costs, the development of AI gathered momentum. In the past two decades, AI has been the central focus of the technology industry.
“In today’s world everyone is either using AI or impacted by AI, it is now the core aspect of businesses. It has started to give meaningful and economic results recently towards real world problems. AI is a transformational force right now, similar to computers in the 1970s and 1980s. It is allowing us to deliver the results and helping us create differentiating value faster, better, and affordable way at scale in a highly personalized manner.”
Several industries have genuinely benefited from the rapid development in the AI domain. Parmar comments, “The Pharmaceutical sector which was traditionally seen as laggards in terms of adopting digital technology have come to the forefront of digital transformation due to the pandemic. They are among the fastest adopters of the AI technology along with other high-tech industries.”
A similar story is being repeated across different industrial sectors and their clients. Everyone on an individual basis is now benefitting from the development in the AI domain.
The reason why AI needs to be secured
AI models require a large amount of sensitive training data and are usually computationally intensive to build. The commercial value of such models motivates attackers to mount various attacks.
Adversaries can launch attacks for monetization purposes or as a stepping-stone towards other attacks like model evasion. In specific cases, it even results in destroying brand reputation, differentiation, and value proposition.
The fear of security of AI is one of the most important factors hindering the speed of its adoption. AI Security is currently an emerging theme and will become a vital aspect to continue adopting Global AI transformation and positively impacting human life, society, and the environment.
In today’s world when we are relying on AI algorithms to detect and respond to cyber-attacks, it is of utmost importance to secure those algorithms. The increased dependence on AI by industries across the world will instigate attackers to target these algorithms, which can lead to severe consequences.
Adversaries may like to take hold of the AI systems or indulge in data poisoning to confuse the algorithms. With a paradigm shift to AI dependence following the Coronavirus outbreak in the world, its security is of paramount importance.
AI attacks can (and are getting) more dangerous in the future - rather than just hacking your laptop, if given access to the wrong hands, AI can have a negative impact of global proportions. AI attacks such as Deepfake and Smart Phishing attacks are on the rise.
“AI must be protected against emerging security threats and specifically against vicious model extraction attacks. Successful model extraction attack, that aims to copy the AI model, nullifies the entire storyline of ‘Data is the new oil’. If AI is not protected, all the investments done to develop it will be put at significant risk and in extreme cases can be of little financial worth.”
The cybersecurity threat to Artificial Intelligence is higher than ever as per several surveys and AI index reports.
Parmar points out, “The cybersecurity threats were not given enough emphasis when the era of computers was at its peak. This has now left us with huge vulnerabilities and issues that we are still grappling to fix. The story has repeated with mobile and IOT era as well. The cybersecurity breaches and the losses caused due to it are running into billions of dollars in today’s date.”
“Security remains one of the biggest barriers for the adoption of AI in the products as well as internal adoption across industries.”
Steps being taken to secure AI
To counter threats for AI, industry players have developed an open-source framework called MITRE.ATLAS (Adversarial Threat Landscape for Artificial Intelligence Systems) in which Bosch is a co-founder.
“We consistently keep an eye for all the emerging threats which are available in the AI domain, we have tried to categorize all of them and put them into perspective for the wider community to understand it and spread awareness about it. It will prepare organize to tackle the emerging threats.”
There are a lot of cases of security threats that are present and the capacity to fool existing AI systems.
“At Bosch, we are currently building a product called – AI Shield (currently in testing stage) which will look into an AI model and find what kind of vulnerability is associated with the model and will also suggest real-time defense mechanism to provide defense against the threat.
Though cyber-security experts can understand about the vulnerability of the product, they are very scarce in number for traditional field. The defense security experts are in high demand and charge a high cost to detect and suggest a defense mechanism for the existing threat in traditional cybersecurity domain. In emerging AI security field, there are only handful of experts across the world. Again, scarcity of expertise at intersection of AI and security, makes it prohibitive to engage defense experts to generate defenses.”
The dark side of AI adoption
Parmar is of the view that if we take a look at any technology innovation through a crystal ball, it is true that all of them have a positive and a negative side attached to it. AI is no different. As Parmar says, “Even the benefactor of the Noble Peace Prize - Alfred Nobel was inventor of dynamite!”
Technology remains a two-sided sword. He admits, “Everyone is aware that AI can be used for nefarious things such as fraud, ransom attacks etc.”
Related blog - Dark Side of IT
However, Parmar remains optimistic about the role of AI technology in the upcoming future.
“Unlike in the earlier times, people are now openly discussing AI, not only people but organizations and governments are also talking and debating about it. Several organizations and countries are playing an important role in finding out ways to reduce the negative side of the AI technology and focus on the positive aspects.”
The sudden paradigm shift to AI
Industries across sectors have shifted to using AI or are in the process to do so. The year 2020, saw a complete lockdown being initiated in countries across the world. Schools, Colleges and Offices moved online within months. This move, in turn, provided a massive push to propel the shift to AI.
Parmar says that they are highly committed to the development of AI. Bosch CEO, Dr Volkmer Denner has recently announced that by 2025 all Bosch products will either contain AI or will have been manufactured using its help.
“At Bosch, we only preach when we use it first. We have established AI Codex and talking about building trust in AI along with development in AI. The statement to have AI involvement in all of Bosch products along with governing AI codex represents the overarching commitment of technology in AI and developing it in a trustworthy way.”
Bosch has played an important role in coining out the term Industry 4.0 – The digital transformation of traditional manufacturing and industrial practice that has helped businesses improve their production, process, and output.