People Driven and Tech Enabled – How AI and ML are Changing the Future of Cyber Security in India
On the 27th of February, Peter Sparkes the Senior Director, Cyber Security Services, Symantec conducted a webinar on the ‘5 Essentials of Every Next-Gen SOC’. In this webinar, he evaluated the problems that Security Operations Centers (SOCs) are currently facing, and explored possible solutions to these problems. The webinar also put emphasis on AI and ML as tools to improve cyber security. This blog draws key insights from the webinar, and explains how AI and ML can improve the cyber security process of Indian enterprises.
Introduction
In a study conducted by Cisco, it was found that in the past 12-18 months, cyber attacks have caused Indian companies to incur financial damages amounting to USD 500,000.
There is a need to strengthen the nodal agencies in an enterprise that can deal with these threats to prevent irreparable damage to enterprises and their customers. An SOC within any organization is the team responsible for detecting, monitoring, analyzing, communicating and remedying security threats. The SOC technicians employ a combination of technologies and processes to ensure that an enterprise’s security is not compromised. As instances of cyber attacks increase both in number and sophistication, SOCs need to use state of the art technologies to stay one step ahead of the attackers. Presently, SOCs face a number of infrastructural problems such as the low priority given to a cyber security budget, slower and passive response to threats, dearth of skilled technicians, and the absence of a global intelligence network for cyber-threats. This is where technologies such as Artificial Intelligence and Machine learning are helping, by monitoring the system to identify cyber attacks, and analyse the severity of the threat, and in some cases by blocking such threats.
Evolution of Security Operations Centers
In the same study, Cisco looked at the evolution of cyber threats and how companies were using technologies such as AI and ML to ameliorate those threats. Another key insight the study brought out was that 53 and 51 percent of the subject companies were reliant on ML and AI respectively. One of the reasons behind AI and ML’s effectiveness in cyber security is their capacity not only to detect known threats but also to use their learnings from data to detect unknown threats. In his webinar, Peter Sparkes also stated that SOCs were evolving into a ‘people driven and tech enabled’ system.
People Driven and Tech Enabled
In the case of cyber security, which in itself is a relatively new field, technologies such as AI and ML are helping companies to not only overcome infrastructural barriers but also to respond proactively to threats. A study conducted by the Enterprise Strategy Group, revealed that one-third of the respondents believed that ML technology could detect new and unknown malware.
The study also stated that the use of machine learning to detect and prevent threats from unknown malware reduced the number of cases the cyber security team had to investigate.
Similarly, the tasks of monitoring and blocking which were earlier conducted by entry level analysts were now done by systems, using machine learning. Typically, the AI acts as the first monitoring system after which the threat is examined by the company’s technicians who possess the requisite skill set and experience. By delegating the time consuming task of continuous monitoring to an ML system, the technicians now have time to look at serious threats. In this way AI and humans are working together to build a stronger and responsive security protocol.
Detecting the Unknown
Cyber criminals are becoming increasingly sophisticated, and in order to prevent attacks the monitoring systems (both human and automated) need to be able to detect them before the security is compromised. The detection of threats through AI and ML is done in a similar way as it is done for the identification of spam, where the system is trained on a large amount of data which teaches the algorithm to identify right from wrong.
There have been numerous cases of stealthy cyber attacks such as wannacry and ransomware, that have evaded detection by conventional security firewalls and caused crippling damage. There is also the need to use deception technology which involves automatic detection and analysis of attacks. This technology then tricks the attackers and defeats them to bring back normalcy to the system.
The systems that can handle threats by themselves do so by following a predetermined procedure, or playbook where the AI detects activities that go against the procedure/playbook. This is more effective compared to the earlier system where the technicians would analyse the attacks on a case by case basis.
AI and ML can help in reducing the time required to detect threats enabling technicians to act proactively and prevent damage. As AI and ML systems are less prone to make mistakes compared to human beings, each threat is dealt with in a prompt and accurate manner. AI systems also help by categorising attacks based on their propensity for damage. These systems can use the large volumes of data collected about previous attacks and adapt over time to give enterprises a strong line of defence against attacks.
Passive to Active Defense
Threat to cyber security can emerge even in seemingly safe departments, such as Human Resources. It is therefore important to proactively hunt for threats across all departments uniformly.
In order to detect an anomaly, the AI and ML system will require both large volumes of data as well as a significant amount of processing power, which is difficult for smaller companies to provide. A possible solution to improve defense is to have a system of sharing SOC data between companies, and thereby creating a global database of intelligence. A system of global intelligence and threat data sharing could help smaller companies combat cyber threats without having to compromise on core business development.
Use of AI in Cyber Security in India
In 2017, Indian enterprises were infected by two lethal cyber attacks called Nyetya that crept through a trusted software - Ccleaner and infected computers
. These attacks may just be the tip of the iceberg , since there may be many other attacks that might have gone unreported, or worse, undetected. Cisco reported that less than 55 per cent of the Indian enterprises were reliant on AI or ML for combating cyber threats. Although the current numbers seem bleak, there are a number of Indian enterprises that have recently begun using AI and ML in cyber security.
One such example is HDFC bank which is in the process of introducing an AI based Cyber Security Operations Centre (CSOC).
This CSOC is based on a four point approach to dealing with threats - prevent, detect, respond and recover. The government of India has also taken its first step towards the use of AI in cyber security through a project that aims to provide cyber forensic services to the various agencies of the government including law enforcement.
Indian intelligence agencies have also entered into an agreement with tech startup Innefu, which utilizes AI, to process data and decipher threats by looking at the patterns of past threats.
As India is increasingly becoming data dense both private and public organizations need to consider cyber security with utmost seriousness and protect the data from crippling attacks.
Conclusion
Enterprises have become storehouses of user data and the SOCs have a responsibility to protect this data. The companies’ SOCs have been plagued with several problems such as lack of skilled technicians, delay in response time and the inability to proactively respond to attacks. AI and ML can help in a system of continuous monitoring as well as take over the more repetitive and time consuming tasks, leaving the technicians with more time to work on damage control. Although it must be kept in mind that AI is not a silver bullet, since attackers will try their best to confuse the AI systems through evasion techniques such as adversarial AI (where the attackers design machine learning models that are intended to confuse the AI model into making a mistake).
Hence, human intervention and monitoring of AI and ML systems in cyber security is essential to maintain the defence and protection mechanisms of enterprises.
A few topics that Indian SOCs need to consider while using AI and ML :
1. The companies need to understand that AI and ML need human expertise and supervision to be effective and hence substituting people for AI is not ideal.
2. The companies need to give equal if not more importance to data security.
3. The companies need to constantly upgrade their systems and re-skill their technicians to combat cyber security threats.
4. The AI and ML systems need to be regularly audited to ensure that they are not compromised by cyber attacks and also to ensure that they are not generating false positives.
[]. Cisco, (2018, February). Annual Cybersecurity Report. Retrieved March 8, 2018, from https://www.cisco.com/c/dam/m/digital/elq-cmcglobal/witb/acr2018/acr2018final.pdf?dtid=odicdc000016&ccid=cc000160&oid=anrsc005679&ecid=8196&elqTrackId=686210143d34494fa27ff73da9690a5b&elqaid=9452&elqat=2
[]. Enterprise Strategy Group (2017, March ). Top-of-mind Threats and Their Impact on Endpoint Security Decisions. Retrieved March 8, 2018 from https://www.cylance.com/content/dam/cylance/pdfs/reports/ESG-Research-Insights-Report-Summary-Cylance-Oct-2017.pdf
[]. Vorobeychik,Y (2016). Adversarial AI. Retrieved March 8, 2018, from https://www.ijcai.org/Proceedings/16/Papers/609.pdf
[]. Quora. ( 2081, February 15). How Will Artificial Intelligence And Machine Learning Impact Cyber Security? Retrieved March 8, 2018, from https://www.forbes.com/sites/quora/2018/02/15/how-will-artificial-intelligence-and-machine-learning-impact-cyber-security/#569454786147
[]. Sparkes, P. (2018, February 27). The 5 Essentials of Every Next-Gen SOC. Retrieved March 8, 2018, from https://www.brighttalk.com/webcast/13389/303251/the-5-essentials-of-every-next-gen-soc
[]. PTI. ( 2018, February 21).Indian companies lost $500,000 to cyber.Retrieved March 8, 2018, from https://economictimes.indiatimes.com/tech/internet/indian-companies-lost-500000-to-cyber-attacks-in-1-5-years-cisco/articleshow/63019927.cms
[]. Cisco, (2018, February). Annual Cybersecurity Report. Retrieved March 8, 2018, from https://www.cisco.com/c/dam/m/digital/elq-cmcglobal/witb/acr2018/acr2018final.pdf?dtid=odicdc000016&ccid=cc000160&oid=anrsc005679&ecid=8196&elqTrackId=686210143d34494fa27ff73da9690a5b&elqaid=9452&elqat=2
[]. Raval, A. ( 2018,January 30). AI takes cyber security to a new level for HDFC Bank.Retrieved March 8, 2018, from http://computer.expressbpd.com/magazine/ai-takes-cyber-security-to-a-new-level-for-hdfc-bank/23580/
[]. “The Centre for Development of Advanced Computing (C-DAC) under the Ministry of Electronics and Information Technology (MeitY) is working on a project to provide cyber forensic services to law-enforcing and other government and non-government agencies.” Ohri, R. (2018, February 15. Government readies AI-muscled cyber security plan. Retrieved March 8, 2018, from https://economictimes.indiatimes.com/news/politics-and-nation/government-readies-ai-muscled-cyber-security-plan/articleshow/62922403.cms utm_source=contentofinterest&utm_medium=text&utm_campaign=cppst
[]. Chowdhury, P.A. (2017, January 30). Cyber Warfare at large in Southeast Asia, India leverages AI for the same cause Retrieved March 8, 2018, from https://analyticsindiamag.com/cyber-warfare-large-southeast-asia-india-leverages-ai-cause/
[]. Open AI.(2017 February 24). Attacking Machine Learning with Adversarial Examples. Retrieved March 8, 2018, from https://blog.openai.com/adversarial-example-research/