Hyderabad

Out-of-the-box Hyderabad: Staying ahead of cyber threats

Founded by Abhishek Sinha, Yakshanexus is developing AI-powered security solutions designed to keep pace with an increasingly complex global cyber threat landscape

Ipshita Raman

A momentary lapse in judgement is all it takes to compromise personal security and expose serious vulnerabilities. Phishing links, malicious websites and vulnerable applications can turn a simple human error into a significant security incident.

Driven by the belief that security should not depend entirely on human vigilance, but should work continuously and intelligently alongside people, Abhishek Sinha, established Yakshanexus in Hyderabad in 2025. An early-stage, AI-powered cybersecurity company, Yakshanexus is designed to detect, analyse and prevent digital threats before they cause harm.

The company offers two primary tools: Yakshanexus P1, an Android app that protects everyday users from phishing attacks, and Yakshanexus WP1, an autonomous AI security engineer that explores applications, investigates vulnerabilities, validates findings and builds evidence around potential attack paths.

Together, these tools address both ends of the cybersecurity challenge: protecting individuals who use the internet every day and securing the organisations and platforms they depend on.

CE speaks to Abhishek about the inception of Yakshanexus, its contributions to the cybersecurity landscape, the state of cybersecurity locally and globally, and the cost of treating it as an afterthought.

Excerpts
What inspired the founding of Yakshanexus?

For more than four years, I worked as a security analyst and penetration tester, auditing and securing both government and private-sector domains. The work was almost entirely manual, and that was its biggest limitation: a human tester can focus on only one target at a time and can find only what their knowledge and tools allow.

Meanwhile, threats continue to grow, with attackers increasingly using AI to move faster and operate at a scale no individual can match. I realised that defenders needed a system capable of thinking, reasoning, acting, exploiting and validating at machine speed while retaining the judgement of a senior tester. Bridging that gap between human-speed defence and machine-speed, AI-armed attacks is why Yakshanexus exists.

How would you assess Hyderabad and Telangana’s readiness to tackle the increasingly complex threat of cybercrime?

I would assess it very positively. Hyderabad has emerged as one of India’s strongest technology and cybersecurity hubs, with a deep pool of security talent, a growing startup ecosystem and institutions such as the Cybersecurity Centre of Excellence (CCoE) actively driving the sector forward. Telangana has been notably proactive in treating cybersecurity as core infrastructure rather than an afterthought. Its willingness to engage with industry and emerging companies reflects genuine preparedness. The foundation is strong, with skilled professionals, supportive institutions and a government that works with innovators. The task ahead is to keep pace with AI-accelerated threats, and this region is better positioned than most to do so.

What further steps can the Telangana government take to strengthen the state’s cybersecurity framework and enhance its ability to combat cyber threats?

The most valuable step would be to deepen collaboration with homegrown cybersecurity innovators so that state systems can benefit from solutions developed for Indian conditions and threats. I would also encourage continued investment in AI-driven defence, as threats targeting public infrastructure are increasingly AI-accelerated, and traditional tools alone may not be sufficient to keep pace.

Strengthening basic cyber hygiene and awareness among citizens and small businesses would also help close the easiest entry points for attackers, which remain among the most commonly exploited. We have been engaging with state cyber authorities on precisely these issues, demonstrating how P1 and WP1 can help protect both officials and citizens.

Can you discuss the WP1 and P1 tools, and how do they help users stay ahead of cyber threats?

Most security tools rely on matching known signatures or patterns, which means they struggle to identify new and emerging threats. WP1, on the other hand, reasons about an application the way a skilled human tester would. It goes beyond known CVEs (Common Vulnerabilities and Exposures) to identify logic flaws, connect multiple steps and validate its findings across web, API (Application Programming Interface), mobile and source code environments.

P1 brings that same real-time approach to individual users by inspecting links as soon as they arrive. Both products are built to operate at the speed at which attackers now move, helping users stay ahead of threats rather than react after the damage is done.

How does P1 distinguish between a legitimate website and a sophisticated phishing one designed to closely imitate it?

P1 does not depend on a blocklist of known malicious links, because sophisticated phishing attacks can bypass such lists simply by registering new domains. Instead, its machine-learning model evaluates each link using 42 signals, including domain age, URL length and structure, character patterns, and anomalies in subdomains and paths.

Even a near-perfect visual imitation of a legitimate website leaves measurable fingerprints in the way its link is structured and hosted. The model evaluates these signals together to generate a score that classifies a link as safe, unsafe or suspicious. This allows P1 to flag a newly created phishing page it has never encountered before, rather than identifying only those already present on a blocklist.

How does WP1 function as an autonomous AI security engineer? What steps does it take when examining an application for vulnerabilities?

WP1 behaves like a penetration tester rather than a conventional scanner. It begins by profiling the target to understand the application and how it functions, mapping its inputs, endpoints and logic. It then analyses the entire codebase or application instead of inspecting individual files or endpoints in isolation. Based on this assessment, it develops hypotheses about potential weaknesses and actively tests them.

When it identifies a potential issue, WP1 works to validate it, confirming that the vulnerability is real rather than flagging a theoretical match. It also coordinates several specialised agents to examine different aspects simultaneously.

The result is a set of confirmed and prioritised findings. The engine is specifically designed to uncover novel logic flaws that fall outside the known CVE and CWE (Common Weakness Enumeration) catalogues.

How much human oversight is required when WP1 conducts security assessments?

WP1 is designed to conduct the core assessment autonomously. It investigates, tests and validates findings independently, which is precisely what distinguishes it from tools that simply generate a list of potential issues for a human to examine. Human oversight is required at two key stages: authorising the scope of the assessment and reviewing the final report, rather than directing every step of the process. This is deliberate, as the objective is to deliver senior-tester-level reasoning at machine speed without creating additional manual work.

How frequently are your AI models updated to keep pace with evolving attack techniques, and can they adapt to new threats in real time?

Our architecture is deliberately model-independent, allowing WP1 to run on current frontier models and improve as those models evolve, without requiring us to rebuild the engine each time. In addition, we continuously refine its agents, reasoning processes and techniques as new attack methods emerge.

P1’s phishing detection model is similarly fine-tuned to recognise the patterns attackers are actively using. Our goal is to keep our update cycles short in response to the rapidly evolving threat landscape, ensuring that users are protected against what is happening now, rather than what was happening last year.

The recent PHP2Shell vulnerabilities triggered a global cybersecurity scare. What does this incident reveal about vulnerabilities in today’s digital infrastructure that businesses should immediately learn from?

PHP2Shell, a pair of high-impact vulnerabilities, involved two critical flaws in PHP (PHP: Hypertext Preprocessor), the software that powers a significant proportion of the world’s websites. They were uncovered with the help of AI. What made the incident alarming was not just the severity of the vulnerabilities but also the speed at which the situation escalated. Exploitation attempts surged within hours of disclosure, prompting governments and companies around the world to take their websites offline as a precaution.

The real lesson is that the gap between a vulnerability becoming known and being exploited on a large scale has shrunk from weeks to hours, as attackers now have AI at their disposal too. Businesses should treat patching public-facing systems as an incident-response priority rather than routine maintenance. They should also test their applications continuously instead of relying on annual assessments. This is precisely the shift WP1 is designed to support: identifying vulnerabilities at machine speed before attackers can exploit them.

What would you say to individuals and businesses that still consider cybersecurity an afterthought?

Cybersecurity is no longer optional. Almost everything we own and operate is now digital, and an attacker needs only one opening. For individuals, the basics matter: avoid opening unfamiliar links, be wary of offers that sound too good to be true, and never share your card number or PIN to unblock an account or claim a prize you never entered. Classic scams are everywhere, whether it is a message claiming you have been selected from among 1,00,000 people for a reward, a stranger saying they have named you in their will, or a fake bank alert asking you to verify your debit card through a link.

For businesses, one of the most dangerous assumptions is that they are too small to be targeted. Cybersecurity is no longer an expense to defer; it is what keeps everything else standing.

'Ahmedabad Air India crash probe in final stages': Centre tells SC

US Democrats prepare probes into Trump family, administration ahead of taking control of House panel

‘If we can end poverty, we can end drugs’: PM Modi calls for intensified anti-narcotics drive

Odisha Police deploys 18 ODRAF teams across seven districts ahead of Cyclone Akvan

Delhi court seeks police report on complaint over Sonia Gandhi’s name in 1980 electoral roll