Introduction
Zero-day brokering did not emerge as an organised market. Sreeja traces its origins to the early days of computing, when vulnerabilities were often discovered and exploited by curious programmers and hobbyists, not necessarily with any commercial interest. As digital technology developed, the practice took a darker turn: zero-day vulnerabilities came to be exploited by cybercriminals for financial gain and by nation-states for espionage and cyber warfare.1 Dellago, Woods and Simpson date its commercialisation to the mid-2010s, when brokers such as Zerodium began publicly advertising the prices they would pay researchers for exploits, something previously unseen, since market participants had until then navigated informal professional networks.2 Their research illustrates the growth that followed: the mean price of exploits rose by 1,240% over their six-year observation period, and the maximum price quoted by Zerodium reached $2.5 million in 2019.3 Bilge and Dumitraş had earlier put the market value of a new vulnerability at between $5,000 and $250,000.4
Several incidents show the damage a zero-day vulnerability can do once someone who knows of it, while the affected vendor does not, turns it into a weapon. The most prominent example is probably Stuxnet, a worm that exploited four zero-day vulnerabilities in Microsoft Windows to attack the industrial control systems of Iran’s nuclear facilities in 2010, spinning the uranium-enrichment centrifuges until they malfunctioned and damaging some 1,000 of them.5 Another major vulnerability chain is HAFNIUM’s exploitation of four zero-days in Microsoft Exchange Server (CVE-2021-26855, -26857, -26858 and -27065) to achieve unauthenticated remote code execution, which by 5 March 2021 had affected, on estimates, more than 30,000 organisations in the United States alone.6 Finally, the Log4Shell vulnerability (CVE-2021-44228) is a particularly egregious case. A flaw in the way the Log4j logging library handled log messages allowed an attacker to plant a JNDI lookup string that made the library fetch and execute code from a remote server, giving remote code execution on any system running an affected version. The library is built into products from vendors such as Amazon and Cloudflare and into games such as Minecraft. Within 72 hours of the flaw being publicly reported, more than a million exploitation attempts had been recorded, ransomware groups (Conti, LockBit) and botnets (Mirai variants) soon incorporated it, and the vulnerable code had been present in the library, unnoticed and unpatched, since 2013.7 In each of these cases the fundamental problem was the same: a vulnerability unknown to the affected vendor (or, in the case of Log4j, to the maintainers of the library) was turned into an attack vector by someone who knew of it in advance.
Indian law imposes no liability for brokering an undisclosed vulnerability, has no safe harbour distinguishing researchers acting in good faith from malicious actors, and mandates no disclosure timeline. This paper asks where in the lifecycle of a zero-day vulnerability Sections 43, 66, 66F, 70B and 75 of the Information Technology Act, 2000 cease to reach the conduct of brokers, and proposes amendments to close that gap, reading those provisions doctrinally and testing the reading against the Supreme Court’s Pegasus judgment and four foreign models: the US Computer Fraud and Abuse Act, the EU Cyber Resilience Act, Israel’s export-control regime and China’s vulnerability provisions.
The liability gap in Indian law
Zero-day vulnerability brokering is defined nowhere in any of the statutes reviewed. The IT Act defines “access”, “computer”, “computer network” and “computer system”.8 The CFAA defines “protected computer”.9 The EU Cyber Resilience Act defines “actively exploited vulnerability”, but not “zero-day vulnerability”.10 The Chinese Provisions regulate “network product security vulnerabilities”.11 The Wassenaar-derived Israeli export-control regime controls “intrusion software”, not vulnerabilities as such.12 If “zero-day vulnerability” were defined in positive law anywhere, that definition could be borrowed here. Instead, the term exists only in technical literature and policy discussion. The definition must therefore be supplied in the statute itself; once it is, the analysis can move on to applying the statute to that definition.
Sections 43 and 66 require actual, unauthorised access before liability attaches. Brokers who find and evaluate vulnerabilities, enter contracts to sell them, and sell them without ever touching the intended target do not meet that precondition. Bilge and Dumitraş’s finding that a zero-day attack lasts on average 312 days before the vulnerability is disclosed also suggests that this loophole is large enough to matter.13 The activity prohibited by Sections 43 and 66 is intrusion itself; the activity the market literature identifies as “brokering” happens entirely before any intrusion occurs. In that literature a “zero-day vulnerability” is necessarily distinguished from an “exploit”, and both Dellago et al. and Sreeja describe the broker’s role as evaluating, holding and selling vulnerabilities.14 Taken together, these elements yield the following position: a sale occurs without access, so a brokering provision would not overlap with the existing access offences, and selling a zero-day and using it remain distinct acts under these sections.
Properly construing these sections requires a distinction between two timelines the literature conflates: the patch clock and the disclosure clock. Roumani’s evidence shows that patch times vary with a vulnerability’s attributes and with the type of product affected, which argues for flexibility in patching.15 Disclosure faces no comparable barrier: a person who knows of a vulnerability can report it at once. CERT-In already requires specified cyber incidents to be reported within six hours of being noticed,16 and Article 14 of the EU Cyber Resilience Act requires a manufacturer to give early warning of an actively exploited vulnerability within twenty-four hours of becoming aware of it.17 The framework proposed here therefore adopts a graduated reporting duty that runs from awareness: an initial notification to the affected vendor or CERT-In within six hours, a detailed report within seventy-two hours, and a final report within fourteen days of a fix becoming available. Further public disclosure can and should follow, and the proposed framework does not limit how the vulnerability is disclosed after that point.
The central analytical difficulty is that a good-faith researcher and a malicious broker are indistinguishable at the moment of discovery and differ only in what they do next. The narrowing of the CFAA in Van Buren18 illustrates the failure of access-based law to resolve the problem at the access event itself, since both actors access, or discover, the flaw in the same way. China’s Article 4 regime is instructive by negative example: it prohibits the sale of vulnerability information while providing no good-faith mechanism at all, collapsing the distinction between the researcher and the malicious actor into a single undifferentiated prohibition.19 The workable standard therefore looks to conduct after discovery: an identifiable reporting party, compliance with the reporting window, and the absence of the transactional markers (specification sheets, preliminary offers, staggered payments) that the broker literature identifies as characteristic of a sale.20
Because a zero-day’s severity cannot be conventionally scored in advance, the analysis draws on the k-zero-day safety metric, for which Albanese et al. propose efficient estimation algorithms, and which measures an asset’s resilience by how many distinct unknown vulnerabilities would be required to compromise it.21 Critical infrastructure is structurally more fragile on this measure, as a single additional zero-day would materially reduce its resilience, which justifies the enhanced treatment of Section 66F where the broker knew or should have known of the target’s critical status.22
On extraterritoriality, the comparative review of Israel’s export-licensing regime shows that licensing alone regulates only the point of sale, with post-sale conduct and harm lying beyond the licensing state’s jurisdiction. Section 75 is therefore the appropriate complement, extending liability to offshore brokers wherever the vulnerability or the affected system is located in India.23
Literature review
The literature on zero-day vulnerabilities may be grouped into four converging strands: the empirical study of zero-day exploitation in the wild, quantitative models of vendor patching behaviour, the technical and economic anatomy of the vulnerability trade, and the technical assessment of undisclosed risk. Read together, these strands support the case for a conduct-based, discovery-triggered disclosure obligation of the kind proposed in this study; the doctrinal and comparative analysis follows in the next Part.
A. Prevalence, duration and real-world impact of zero-day exploitation
The foundational empirical study in this area is Bilge and Dumitraş (2012), who used binary-reputation and anti-virus telemetry data covering 11 million hosts and some 300 million files to identify zero-day attacks retrospectively.24 Their method identified 18 vulnerabilities exploited before public disclosure, 11 of which had not previously been recognised as zero-days, suggesting that zero-day exploitation is considerably more common than industry reporting (which identified only 8 to 15 such vulnerabilities a year between 2006 and 2011) had suggested.25 Their central finding, that a typical zero-day attack lasts on average 312 days, with some running as long as 30 months, and that public disclosure is followed by an increase in attack volume of up to five orders of magnitude, is the empirical anchor for treating the pre-disclosure window as the period of greatest societal risk.26 The authors caution that their results are best interpreted as lower bounds;27 the true undisclosed period may be considerably longer. This finding undercuts any assumption that market-based non-disclosure is a victimless or purely private transaction: the longer a vulnerability sits undisclosed, brokered rather than reported, the larger the population of ultimately affected systems.
This large-N study is complemented by detailed case analyses of the HAFNIUM campaign against Microsoft Exchange Server (2021) and the Log4Shell vulnerability in the Apache Log4j library (2021), which illustrate the mechanics of exploitation once a zero-day is weaponised. The HAFNIUM intrusion exploited a chain of four vulnerabilities (CVE-2021-26855, -26857, -26858 and -27065), comprising server-side request forgery, insecure deserialisation and arbitrary file-write flaws, and by 5 March 2021 had affected more than 30,000 organisations in the United States alone.28 The Log4Shell flaw, a remote-code-execution defect that had existed undetected in the library since 2013 before its disclosure in 2021, generated over a million exploitation attempts within 72 hours of being publicly reported and was rapidly absorbed into ransomware (Conti, LockBit) and botnet (Mirai-variant) operations.29 Both case studies converge on the point Bilge and Dumitraş make statistically: harm is compounded while a vulnerability is known to someone (a researcher, a broker, an attacker) but not to the vendor or to a body capable of coordinating a fix.30
B. Vendor patching behaviour and the design of disclosure timelines
A second body of work addresses the vendor’s side of the exposure window rather than the attacker’s. Roumani (2021) applies Cox proportional-hazards survival analysis to 4,897 zero-day vulnerabilities disclosed through a major bug-bounty programme, the Zero Day Initiative (ZDI), between 2010 and 2020, modelling patch-release time against vulnerability attributes drawn from the CVSS framework.31 The study suggests that a fixed disclosure deadline shapes vendor behaviour in practice: average patch-release time fell from 193 days in 2012 to 97 days in 2019, during a period in which the programme gave vendors 120 days to patch before disclosure.32 Critically for the design of any statutory timeline, the model shows that not all vulnerabilities behave identically under a single flat deadline: vulnerabilities requiring privileges to exploit, or that primarily threaten confidentiality, were significantly less likely to be patched within the deadline, whereas vulnerabilities whose exploitation causes a “scope change” (affecting resources beyond the vulnerable component) were more likely to be patched on time.33 Subanalyses also show that patch timelines vary systematically by product type (enterprise software has the highest median patch-release time, at 96 days, against 74 days for personal software) and by vulnerability class under the Common Weakness Enumeration taxonomy (a use-after-free defect is patched fastest, at a 71-day median, against 102 days for an incorrect type-conversion defect).34 The practical implication for statutory design is twofold: a fixed external deadline demonstrably compresses the exposure window, but a single undifferentiated clock is likely to be either too lax for easily patched, high-visibility flaws or unrealistically tight for privilege-dependent, confidentiality-impacting ones. That caution favours a graduated reporting structure over a single clock, although it bears more on the later stages of reporting and patching than on the initial notice.
A more recent machine-learning analysis of 415 vulnerability disclosures from January to April 2024 extends this literature into predictive severity modelling.35 Using TF-IDF text features, structured vendor metadata, and classifiers ranging from logistic regression to LSTM and CNN architectures, the study achieves severity-prediction accuracies of about 95%, with keywords such as “privilege escalation” and “remote code execution” emerging as strong textual predictors of high-severity (CVSS ≥ 7.0) classification.36 While the methodological contribution is largely feature engineering and model comparison, the dataset is consistent with Roumani’s evidence that vulnerabilities are not randomly distributed but cluster around identifiable technical and vendor-level features,37 which reinforces the case for a statute that ties reporting obligations to objectively verifiable disclosure-agency data (CVE assignment, CVSS scoring) rather than to a discoverer’s potentially self-interested characterisation of severity.
C. The commercial market for vulnerabilities and the role of brokers
A more directly relevant literature addresses the structure of the market in which discovered but undisclosed vulnerabilities are traded. One systematic account of this grey economy describes vulnerability markets as operating on ordinary supply-and-demand logic: researchers seeking financial reward supply zero-days, while cybercriminals, government agencies and even security vendors constitute demand. This literature catalogues the range of buyer motivations (cybercriminal profit, state espionage and surveillance, offensive military capability, defensive acquisition) and identifies brokers as intermediaries who connect researchers to buyers, often allowing both sides to remain pseudonymous and thereby introducing opacity into the transaction.38 This analysis is significant for statutory design in two respects. First, it documents that responsible disclosure is not merely a technical process but is beset by genuine ethical dilemmas for researchers, who face the lucrative pull of vulnerability markets as an alternative to working with vendors.39 Second, it notes that legal and regulatory responses to this market are fragmented and jurisdiction-specific, with export-control regimes proving difficult to enforce in the absence of a global regulatory framework,40 a gap the extraterritorial-reach proposal in this study directly targets.
A related review situates this market within its cross-industry consequences, cataloguing the 2020 Apple iOS zero-click compromises, the 2016 SWIFT banking-network intrusion, and the WannaCry ransomware’s exploitation of healthcare infrastructure as evidence that the risk of an undisclosed, tradeable exploit is not confined to any one sector.41 Drawing on industry risk-analysis data, this review reports that remote code execution vulnerabilities comprised 17% of vulnerabilities in one reporting period, while financial institutions alone accounted for 30% of recorded zero-day incidents.42 It also records a finding of particular relevance to the timeline debate: that organisations took forty-five days on average to address a vulnerability after its disclosure, a critical window of exposure attributed to insufficiently urgent communication between researchers, vendors and industry stakeholders,43 which is precisely the coordination failure a mandatory, short reporting window is designed to correct.
D. Technical assessment of undisclosed risk
A distinct, more formal strand of the literature addresses the foundational difficulty that a zero-day vulnerability carries, by definition, none of the vendor-assigned severity data on which conventional scoring systems such as CVSS depend. One technical contribution builds on the “k-zero-day safety” metric, which measures a network asset’s resilience by the number of distinct, hypothetical zero-day vulnerabilities an attacker would need to chain together to compromise it (a higher number indicating greater structural security, whatever any individual unknown flaw turns out to be), and proposes a set of polynomial-time algorithms that estimate the metric without requiring a complete attack graph to be generated in advance, which is often computationally infeasible for realistically sized networks.44 For present purposes, the significance of this technical literature is inferential rather than doctrinal: because an unknown vulnerability’s risk cannot be scored in the way a known one can, an effective legal response is better designed to attach liability to the conduct of discovering, holding and trading the flaw without disclosure than to any post hoc assessment of how severe the flaw ultimately proved to be. The same logic supports treating vulnerabilities affecting critical infrastructure (where a single additional zero-day materially reduces the resilience of an essential asset) as warranting the heightened statutory tier proposed under Section 66F,45 rather than folding them into an undifferentiated liability standard.
Analysis and comparative perspective
A. The Indian statutory and regulatory framework
Turning to doctrine, the architecture of Indian cyber law shows the structural gap this study addresses. Section 43 of the Information Technology Act, 2000 attaches liability to specified acts done “without permission of the owner or any other person who is incharge of a computer, computer system or computer network”, and Section 66 makes the same acts an offence when they are done “dishonestly or fraudulently”;46 both provisions are aimed at the act of intrusion rather than at what a person does with knowledge of a flaw discovered through lawful means such as independent research. Section 70B already makes CERT-In the national agency for incident response, empowered to issue guidelines, advisories and vulnerability notes and to give binding directions,47 but India has no statutory safe harbour for good-faith security researchers, a gap the proposed carve-out for registered researchers who report within the prescribed window is designed to fill. The Digital Personal Data Protection Act, 2023 imposes breach-notification duties only once a personal data breach has occurred,48 leaving unaddressed the upstream question this study targets: what obligation, if any, attaches to a person who discovers an unexploited vulnerability but has not yet caused or suffered a breach.
B. Judicial treatment of alleged exploit-based surveillance
The Supreme Court’s judgment in the Pegasus spyware matter49 supplies the most directly relevant judicial authority, addressing alleged surveillance rather than the vulnerability trade itself, but doing so on allegations inseparable from that trade. The Court recorded reports that the Pegasus spyware could compromise devices through zero-click vulnerabilities, precisely the kind of exploit catalogued in the technical literature above, and that numbers belonging to Indian journalists, doctors, political figures and even some court staff appeared on a leaked list of suspected targets; it also recorded that the Union of India, despite repeated opportunity, placed on record only a “limited affidavit” that did not shed any light on its stand.50 The Court held that surveillance by the State must be justified on constitutional grounds and be necessary, proportionate and evidence-based, that indiscriminate spying on individuals cannot be allowed except with sufficient statutory safeguards, and that a bare invocation of “national security” does not immunise the State from judicial scrutiny: the State must specifically plead and prove on affidavit the grounds said to justify withholding information rather than rely on the mention of national security as a categorical shield.51 Finding the omnibus and vague denial in the Union’s “limited affidavit” insufficient, the Court constituted an independent Technical Committee, functioning under the oversight of a former Supreme Court judge, to investigate the allegations.52
C. Scale and opacity of zero-day harm
The harm caused by zero-day vulnerabilities presents a scale-and-opacity problem: its true extent is almost certainly greater than any confirmed number suggests (scale), and it persists within a brokering market structured, through pseudonymous dealing and jurisdictional fragmentation, to avoid generating visible records in the first place (opacity). Table 1 sets out the figures the sources reviewed in this study report.
| Year | Incident or study | Geographic scope | Affected count (as reported) | Source |
|---|---|---|---|---|
| 2008–2011 | Zero-day attacks identified through WINE telemetry | Global (11 million hosts; no country breakdown) | 18 zero-day vulnerabilities identified; hosts targeted per vulnerability ranged from 1 to about 1.5 million (CVE-2010-2568, one of the vulnerabilities exploited by Stuxnet) | Bilge & Dumitraş (2012)53 |
| 2018 | Citizen Lab report on Pegasus spyware | Nearly 45 countries (global) | Individuals in nearly 45 countries suspected to have been affected; no breakdown by country | Pegasus judgment54 |
| 2020 | Citizen Lab–Amnesty International investigation of a spyware campaign | India (targeted subset) | 9 individuals in India allegedly targeted | Pegasus judgment55 |
| 2021 | Microsoft Exchange Server (HAFNIUM) | United States (primary reporting); global | More than 30,000 organisations affected in the United States alone, and hundreds of thousands globally, by 5 March 2021; no India-specific figure | Waheed et al. (2024)56 |
| 2021 | Pegasus consortium investigation (nearly 17 media organisations) | India (subset of a larger global list) | Nearly 300 numbers belonging to Indians identified on a leaked list; nearly 10 Indian devices allegedly forensically analysed | Pegasus judgment57 |
| 2023 (Nov.) | Microsoft security updates; CrowdStrike risk analysis | Global (vendor level) | 58 vulnerabilities patched, including 5 zero-days (3 under active exploitation); no country-wise or person-level data | Samuel Prakash (2023)58 |
| 2024 (Jan.–Apr.) | ZDI vulnerability disclosures | Global (vendor level, not country level) | 415 vulnerabilities disclosed; no per-country or per-person affected count | Shet & Alsmadi (2025)59 |
Table 1: Reported scale of zero-day exploitation and spyware incidents
A pattern worth noting in the literature reviewed above is the near-total absence of person-level, country-disaggregated reporting on zero-day harm. The empirical field studies measure scale in terms of hosts infected or organisations compromised, millions of hosts in one dataset and tens of thousands of organisations in another,60 while the Indian material available is confined to the fragmentary, litigation-driven figures that emerged in the Pegasus proceedings: a report of individuals in nearly forty-five countries, a campaign that allegedly targeted nine individuals in India, and a later list on which nearly three hundred numbers belonged to Indians, of which about ten devices were allegedly forensically analysed.61 No source reviewed in this study reports a standardised, yearly count of affected individuals or entities in India comparable to the aggregate global figures the technical and empirical literature otherwise produces. This gap is not incidental. It is a direct consequence of the absence of any mandatory reporting channel through which such figures could be systematically collected: the numbers that do exist for India surfaced only because independent investigative consortia and forensic experts pieced them together outside any statutory process, and even then the Union of India did not place corroborating figures on record before the Supreme Court.62 This absence of routine, verifiable disclosure data is itself evidence for the reporting-timeline proposal advanced in this study. A discovery-triggered obligation to notify CERT-In within the six-hour window proposed under an amended Section 70B(6)63 would not only shorten the exposure window a vulnerability enjoys before a fix is possible, but would for the first time generate the kind of systematic, entity-level Indian dataset that presently does not exist, one that could in future support exactly the empirical scrutiny this study has had to construct indirectly from litigation records and from global datasets never designed to answer an India-specific question.
D. Comparative note: authorisation-based liability, mandatory disclosure timelines and export-control regimes abroad
1. The US Computer Fraud and Abuse Act, an authorisation-based model, and its limits for brokering. The CFAA, 18 U.S.C. § 1030,64 criminalises access to a “protected computer” that is either “without authorization” or “exceeds authorized access”. The CFAA is therefore an access-based offence, its reach determined by the accessor’s entitlement to the information on the computer. In this respect it is directly comparable to Sections 43 and 66 of the IT Act: the offence is triggered by accessing information one is not entitled to access. In Van Buren v. United States, however, the US Supreme Court read “exceeds authorized access” narrowly.65 It held that a person exceeds authorised access when he accesses a computer with authorisation but then obtains information located in particular areas of the computer, such as files, folders or databases, that are off limits to him, and it rejected the broader reading under which using a computer for an unauthorised purpose would itself exceed authorised access. A person who is entitled to access information and later puts it to an improper use therefore falls outside the CFAA’s criminal prohibitions. A researcher who discovers a vulnerability through authorised testing, code review or a bug bounty programme and then sells it to a third party would not, on this reading, exceed authorised access. The CFAA thus reaches the act of access, not the later sale: it cannot distinguish the researcher who reports a flaw from the one who sells it, because both obtained the information in the same way. In practice, however, the United States does have a decentralised safe harbour for such research in the industry’s bug bounty programmes and in the Department of Justice’s charging policy for good-faith security research,66 which operate as voluntary limits on the CFAA’s reach. The CFAA experience therefore supports this study’s core recommendation: that “brokering” must be distinguished from “unauthorised access” and made a distinct ground of liability, civil under Section 43 and, where done dishonestly or fraudulently, criminal under Section 66, because a general authorisation-based offence such as the CFAA or Sections 43 and 66 of the IT Act cannot be stretched to reach disclosure-based brokering.
2. The EU Cyber Resilience Act, a live precedent for a short, awareness-triggered reporting clock. The CRA, Regulation (EU) 2024/2847,67 offers the most direct comparison with the short, awareness-triggered reporting clock this study proposes. Article 14 requires a manufacturer to notify any actively exploited vulnerability in its product to the CSIRT designated as coordinator and to ENISA: an early warning within 24 hours of becoming aware of it, a vulnerability notification within 72 hours, and a final report no later than 14 days after a corrective or mitigating measure becomes available.68 The obligation applies from 11 September 2026, with notifications filed through ENISA’s single reporting platform,69 and it reaches manufacturers located outside the EU whose products are made available on the Union market.70 Article 14 binds only manufacturers, and only in respect of vulnerabilities that are already being actively exploited; it imposes nothing on independent researchers or brokers. It nonetheless supports a discovery-triggered duty on them, for two reasons. First, its clock runs from awareness of the vulnerability rather than from an incident, which is closer to the discovery-based duty proposed here than CERT-In’s current six-hour deadline for incident-based notification;71 the researcher or broker who holds an unpatched vulnerability is the first party aware of it, and the logic of an awareness trigger applies to that party with greater force. Second, the CRA graduates its timeline by the level of detail required: 24 hours for an early warning, 72 hours for a fuller notification, and 14 days for a final report. India could adopt the same structure behind an initial six-hour notice, which would meet the concern Roumani (2021) raises about a uniform, fixed-length window.72 The CRA’s reporting duty was itself controversial: an open letter signed by experts and organisations including Google, the Electronic Frontier Foundation and Rapid7 objected that it would give government agencies knowledge of unmitigated vulnerabilities, open to misuse for intelligence and surveillance, and could discourage good-faith research and disrupt the coordinated disclosure periods, such as the 120 days ZDI gives vendors, on which researchers and vendors rely.73 This objection bears directly on a researcher-facing duty. It supports confining the initial notice to a brief report to CERT-In or the affected vendor, reserving technical detail for the later reports, and pairing the duty with the good-faith exception.
3. Israel’s export-control regime, the paradigm case of a high-brokering jurisdiction regulating after the fact. Israel is the jurisdiction most directly associated with the global commercial zero-day and spyware brokering ecosystem, and its policies are informed by experience in regulating that sector. Exports of “offensive cyber technologies” from Israel are heavily restricted, with a licence required from the Defense Export Controls Agency (DECA) of the Ministry of Defense.74 While Israel is not a Wassenaar Arrangement participant, its 2007 export-control framework is effectively equivalent to it. Following the Pegasus revelations of 2021 (the consortium reports on which the petitioners in Manohar Lal Sharma relied),75 Israel required countries purchasing its cyber technology to sign a declaration restricting their use to “the investigation and prevention of terrorist acts and serious crimes only”.76 The Ministry of Defense also cut the list of countries approved for cyber exports from 102 to 37, dropping, among others, Saudi Arabia and the UAE.77 Israel’s export-control experience bears on this study’s recommendations in two respects. First, it shows that end-use licensing of offensive cybersecurity technologies is a viable regulatory instrument for the central problem of state-on-state brokering (purchases by foreign governments). This supplements rather than replaces the researcher-side safe harbour and civil liability regime this study proposes. Second, Israel’s experience shows the limitations of the export-control approach, chiefly its reactive nature. By design, the DECA licensing regime regulates only the sale, not the use of the technology abroad. Once a foreign government purchases offensive cybersecurity tools, their subsequent use is up to that government, and, as Israel’s Defense Ministry has itself stated, it can respond only after the fact: “In cases where exported items are used in violation of export licenses or end use certificates, appropriate measures are taken.”78 Israel’s experience with the NSO Group, Candiru and other private defence contractors therefore informs this study’s recommendations on Section 7579 and extraterritorial enforcement: to regulate the use of a product sold abroad, a jurisdiction must have extraterritorial enforcement capability. While a licensing regime may be sufficient to regulate direct state-on-state brokering (a licensed vendor selling directly to a foreign government), it is entirely ineffective against sales through private third parties (the same vendor selling spyware to a private contractor abroad).
4. China’s Provisions on the Management of Network Product Security Vulnerabilities, a cautionary counter-model. China is the only jurisdiction reviewed that has enacted a standalone prohibition on selling vulnerability information (Art. 4), together with a two-day window within which network product providers must report vulnerabilities in their products to the Ministry of Industry and Information Technology (MIIT) (Art. 7(2)), a ban on providing undisclosed vulnerability information to overseas parties other than the product provider (Art. 9(7)), and criminal liability under Article 15 for exploiting vulnerabilities or helping others to do so.80 Structurally, this is the closest analogue to the brokering-specific provision this study proposes for Section 43. The regime’s design, channelling undisclosed vulnerability information towards a state ministry while forbidding its disclosure abroad, has attracted documented criticism for turning a disclosure statute into a state vulnerability-stockpiling mechanism.81 The MIIT gains privileged early access, with no safe harbour for independent researchers and no public-interest disclosure channel. The regime is instructive by negative example: the proposed amendments must avoid replicating China’s asymmetric design by coupling the CERT-In reporting channel with the good-faith researcher carve-out and with the transparency-reporting obligation proposed below, so that the channel does not function in practice as a one-way state stockpiling channel with no corresponding accountability.
Suggestions
Section 43 (civil liability): Insert a provision under which any person who knowingly holds, acquires, sells, purchases, transfers or otherwise brokers a zero-day vulnerability without reporting it to the affected vendor or CERT-In within the prescribed window, and thereby causes damage to the affected entity, is liable to pay damages by way of compensation to that entity.
Statutory definitions: Insert the definitions of “zero-day vulnerability” and “brokering” immediately before any liability provision, so that liability is tethered to the language of the definitions and not left to judicial interpretation.
Good-faith carve-out: Make a disclosure by a registered security researcher in good faith within the reporting window an exception to liability under Sections 43 and 66, distinguishing such a researcher from a malicious actor.
Reporting channel and timeline: Require any person who becomes aware of a zero-day vulnerability to report it to CERT-In (under Section 70B) or to the affected vendor in three stages, modelled on the CERT-In Directions issued under Section 70B(6) and on Article 14 of the EU Cyber Resilience Act: an initial notification within six hours of becoming aware of the vulnerability, a detailed report within seventy-two hours, and a final report within fourteen days of a fix becoming available.
Criminal liability under Section 66: Apply Section 66 to make it an offence to broker a vulnerability knowingly and for the purpose of selling it to an unauthorised third party, where this is done with the mens rea Section 66 already requires, “dishonestly or fraudulently”.
Aggravated offence under Section 66F: Specify that brokering a vulnerability affecting critical infrastructure is an aggravated offence under Section 66F, linking it to the cyberterrorism framework.
Extraterritorial reach under Section 75: Extend the application of the IT Act to zero-day vulnerability brokering conducted outside India where either the vulnerability or the victim is located within India, since brokers may maintain legal entities in several jurisdictions to deal with export regulations.82
Institutional follow-through: Instruct CERT-In to publish periodic transparency reports on the disclosures it receives under the new regime, both to meet the norm of governmental transparency articulated by the Supreme Court in the Pegasus judgment83 and to facilitate evidence collection for later empirical research on the effects of the amendments.
Conclusion
Indian law reaches zero-day vulnerabilities only at the point of intrusion: Sections 43 and 66 punish unauthorised access to a computer resource while remaining silent on the sale of a lawfully found flaw; the law provides no safe harbour for those who act in good faith; and it imposes no timeline for disclosure at the point of discovery. Empirical literature has highlighted the costs of this approach, with vulnerabilities lying undisclosed for an average of 312 days,84 while Manohar Lal Sharma highlights the constitutional significance of this market in India.85 No foreign regime fully addresses the issue: the CFAA cannot reach post-access conduct such as a sale, the EU CRA focuses on timing rather than on the transaction itself, and China’s prohibition contains no safe harbour. The eight-point amendment framework proposed above, which would harmonise definitions, impose civil and criminal liability, carve out good-faith actors, impose a reporting window at the point of discovery and extend extraterritorially, represents a solution that no other jurisdiction has adopted in toto. Addressing this gap is a task Indian cyber law can no longer postpone.
*****
Footnotes
1. Sreeja P., Securing Cyberspace: Navigating Zero-Day Vulnerabilities through Discovery, Disclosure Strategies, and Defence Mechanisms, 185(45) Int’l J. Comput. Applications 12, 12 (2023), https://doi.org/10.5120/ijca2023923262.
2. M. Dellago, D.W. Woods & A.C. Simpson, Characterising 0-Day Exploit Brokers 3–4, 6 (21st Workshop on the Economics of Information Security (WEIS 2022), Tulsa, 2022), https://weis2022.econinfosec.org/wp-content/uploads/sites/10/2022/06/weis22-dellago.pdf.
3. Dellago, Woods & Simpson, supra note 2, at 2, 8, 19.
4. Leyla Bilge & Tudor Dumitraş, Before We Knew It: An Empirical Study of Zero-Day Attacks in the Real World, in Proceedings of the 2012 ACM Conference on Computer and Communications Security (CCS ’12) 833, 833 (2012), https://doi.org/10.1145/2382196.2382284 (“the market value of a new vulnerability ranges between $5,000–$250,000”).
5. A. Waheed, B. Seegolam, M.F. Jowaheer, C.L.X. Sze, E.T.F. Hua & S.R. Sindiramutty, Zero-Day Exploits in Cybersecurity: Case Studies and Countermeasure 2 (Preprints.org, Preprint No. 2024072338, 2024), https://doi.org/10.20944/preprints202407.2338.v1; Bilge & Dumitraş, supra note 4, at 833; Danita Samuel Prakash, Zero-Day Vulnerabilities: An In-Depth Analysis 1, 3 (Bournemouth Univ., Preprint, 2023), https://doi.org/10.13140/RG.2.2.12775.01445.
6. Waheed et al., supra note 5, at 5–6.
7. Waheed et al., supra note 5, at 7–9.
8. The Information Technology Act, 2000, § 2(1)(a), (i), (j), (l), No. 21, Acts of Parliament, 2000 (India) [hereinafter IT Act] (defining “access”, “computer”, “computer network” and “computer system”).
9. Computer Fraud and Abuse Act, 18 U.S.C. § 1030(e)(2) (2018) (defining “protected computer”).
10. Regulation (EU) 2024/2847 of the European Parliament and of the Council of 23 October 2024 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act), art. 3(40)–(42), OJ L, 2024/2847, 20.11.2024 (EU) [hereinafter Cyber Resilience Act] (defining “vulnerability”, “exploitable vulnerability” and “actively exploited vulnerability”).
11. Provisions on the Management of Network Product Security Vulnerabilities, Gong Xin Bu Lian Wang An [2021] No. 66 (issued by the Ministry of Industry and Information Technology, the Cyberspace Administration of China and the Ministry of Public Security on July 12, 2021; effective Sept. 1, 2021) (China) [hereinafter China Vulnerability Provisions].
12. Garrett Hinck, Wassenaar Export Controls on Surveillance Tools: New Exemptions for Vulnerability Research, Lawfare (Jan. 5, 2018) (the Wassenaar controls cover “intrusion software” and related systems and technology, not vulnerabilities as such), https://www.lawfaremedia.org/article/wassenaar-export-controls-surveillance-tools-new-exemptions-vulnerability-research.
13. Bilge & Dumitraş, supra note 4, at 839.
14. Dellago, Woods & Simpson, supra note 2, at 3–4; Sreeja, supra note 1, at 15.
15. Yaman Roumani, Patching Zero-Day Vulnerabilities: An Empirical Analysis, 7 J. Cybersecurity tyab023, at 6–8, 10 (2021), https://doi.org/10.1093/cybsec/tyab023.
16. Indian Computer Emergency Response Team (CERT-In), Directions under Sub-section (6) of Section 70B of the Information Technology Act, 2000 Relating to Information Security Practices, Procedure, Prevention, Response and Reporting of Cyber Incidents for Safe & Trusted Internet, No. 20(3)/2022-CERT-In (Apr. 28, 2022), direction (ii) (India) [hereinafter CERT-In Directions] (cyber incidents listed in Annexure I to be reported within 6 hours of noticing them).
17. Cyber Resilience Act, supra note 10, art. 14(2)(a) (early warning within 24 hours of the manufacturer becoming aware of an actively exploited vulnerability).
18. Van Buren v. United States, 593 U.S. 374, 396 (2021) (holding that a person “exceeds authorized access” when he accesses a computer with authorization but then obtains information located in particular areas of the computer, such as files, folders or databases, that are off limits to him).
19. China Vulnerability Provisions, supra note 11, art. 4 (no organisation or individual may illegally collect, sell or publish network product security vulnerability information).
20. Dellago, Woods & Simpson, supra note 2, at 3–4 (specification sheet, non-binding preliminary offer, and payment spread over a few months to a year); Bilge & Dumitraş, supra note 4, at 843 (some fees for new exploits paid in instalments, each depending on the lack of a patch).
21. Massimiliano Albanese, Sushil Jajodia, Anoop Singhal & Lingyu Wang, An Efficient Approach to Assessing the Risk of Zero-Day Vulnerabilities, in Proceedings of the 10th International Conference on Security and Cryptography (SECRYPT 2013) 207, 207 (2013), https://doi.org/10.5220/0004530602070218.
22. IT Act, supra note 8, §§ 66F(1)(A), 70 (cyber terrorism includes conduct likely to adversely affect critical information infrastructure).
23. IT Act, supra note 8, § 75(1)–(2) (the Act applies to an offence or contravention committed outside India if it involves a computer, computer system or computer network located in India).
24. Bilge & Dumitraş, supra note 4, at 834, 838.
25. Bilge & Dumitraş, supra note 4, at 833, 836.
26. Bilge & Dumitraş, supra note 4, at 833–34, 839, 841.
27. Bilge & Dumitraş, supra note 4, at 838, 842.
28. Waheed et al., supra note 5, at 5–6.
29. Waheed et al., supra note 5, at 7–9.
30. Bilge & Dumitraş, supra note 4, at 833, 842.
31. Roumani, supra note 15, at 2–4.
32. Roumani, supra note 15, at 4 (Table 1: average of 193 days to patch release in 2012 against 97 days in 2019; ZDI gives vendors 120 days to release a patch).
33. Roumani, supra note 15, at 6, 10.
34. Roumani, supra note 15, at 6–8.
35. Apurva M. Shet & Izzat Alsmadi, An Empirical Analysis of Zero-Day Vulnerabilities Disclosed by the Zero Day Initiative (January–April 2024) 1–2 (arXiv, Preprint No. 2512.15803, 2025), https://doi.org/10.48550/arXiv.2512.15803.
36. Shet & Alsmadi, supra note 35, at 6, 10, 12.
37. Shet & Alsmadi, supra note 35, at 14.
38. Sreeja, supra note 1, at 15.
39. Sreeja, supra note 1, at 14–15.
40. Sreeja, supra note 1, at 15.
41. Samuel Prakash, supra note 5, at 1–3.
42. Samuel Prakash, supra note 5, at 4.
43. Samuel Prakash, supra note 5, at 4.
44. Albanese et al., supra note 21, at 207–08 (proposing polynomial algorithms to estimate the k-zero-day safety metric first defined by Wang et al.).
45. IT Act, supra note 8, § 66F (cyber terrorism, including conduct likely to adversely affect the critical information infrastructure specified under § 70).
46. IT Act, supra note 8, §§ 43, 66.
47. IT Act, supra note 8, § 70B(1), (4)(e), (6) (CERT-In to serve as the national agency for incident response, to issue guidelines, advisories and vulnerability notes, and to give directions).
48. The Digital Personal Data Protection Act, 2023, § 8(6), No. 22, Acts of Parliament, 2023 (India) (duty to intimate the Data Protection Board and each affected Data Principal of a personal data breach).
49. Manohar Lal Sharma v. Union of India, 2021 SCC OnLine SC 985 (India) (Pegasus spyware case; order of Oct. 27, 2021, N.V. Ramana C.J., Surya Kant and Hima Kohli JJ.).
50. Manohar Lal Sharma, supra note 49, ¶¶ 3, 6, 45 (recording reports that Pegasus could compromise devices through zero-click vulnerabilities and that nearly 300 of the leaked numbers belonged to Indians, and that the Union’s “limited affidavit” did not shed any light on its stand).
51. Manohar Lal Sharma, supra note 49, ¶¶ 34–36, 49–50 (surveillance must be justified on constitutional grounds and be necessary, proportionate and evidence-based; indiscriminate spying on individuals cannot be allowed except with sufficient statutory safeguards; the mere invocation of national security does not render the Court a mute spectator, and the State must plead and prove the need for secrecy on affidavit).
52. Manohar Lal Sharma, supra note 49, ¶¶ 51, 55, 59 (finding the omnibus and vague denial in the “limited affidavit” insufficient, and constituting a three-member Technical Committee overseen by Justice R.V. Raveendran, former Judge of the Supreme Court).
53. Bilge & Dumitraş, supra note 4, at 833, 838–39 (Table 2).
54. Manohar Lal Sharma, supra note 49, ¶ 2 (recording the September 2018 Citizen Lab report that individuals from nearly 45 countries were suspected to have been affected).
55. Manohar Lal Sharma, supra note 49, ¶¶ 4–5 (recording that in 2019 the devices of certain Indians were reported to have been affected, and that in June 2020 Citizen Lab and Amnesty International uncovered a campaign that allegedly targeted nine individuals in India).
56. Waheed et al., supra note 5, at 5.
57. Manohar Lal Sharma, supra note 49, ¶ 6 (recording the July 2021 findings of a consortium of nearly 17 journalistic organisations that nearly 300 of the leaked numbers belonged to Indians and that nearly 10 Indians’ devices were allegedly forensically analysed).
58. Samuel Prakash, supra note 5, at 4.
59. Shet & Alsmadi, supra note 35, at 1–2.
60. Bilge & Dumitraş, supra note 4, at 838–39; Waheed et al., supra note 5, at 5.
61. Manohar Lal Sharma, supra note 49, ¶¶ 2, 5–6 (recounting the 2018 Citizen Lab report (nearly 45 countries), the 2020 Citizen Lab–Amnesty International finding (nine individuals in India) and the 2021 consortium findings (nearly 300 Indian numbers; nearly 10 devices allegedly forensically analysed)).
62. Manohar Lal Sharma, supra note 49, ¶¶ 45, 51.
63. IT Act, supra note 8, § 70B(6) (CERT-In may call for information and give directions to service providers, intermediaries, data centres, body corporate and any other person).
64. Computer Fraud and Abuse Act, 18 U.S.C. § 1030(a)(2), (e)(2), (e)(6) (2018).
65. Van Buren, 593 U.S. at 396.
66. U.S. Department of Justice, Press Release, Department of Justice Announces New Policy for Charging Cases under the Computer Fraud and Abuse Act (May 19, 2022) (directing that good-faith security research should not be charged under the CFAA), https://www.justice.gov/archives/opa/pr/department-justice-announces-new-policy-charging-cases-under-computer-fraud-and-abuse-act.
67. Cyber Resilience Act, supra note 10.
68. Cyber Resilience Act, supra note 10, art. 14(1)–(2).
69. Cyber Resilience Act, supra note 10, arts. 14(1), 16, 71(2) (Article 14 applies from 11 September 2026; the single reporting platform is established by ENISA).
70. Cyber Resilience Act, supra note 10, art. 2(1) (applies to products with digital elements made available on the Union market).
71. CERT-In Directions, supra note 16, direction (ii).
72. Roumani, supra note 15, at 6–8, 10.
73. Michael Hill, Cybersecurity Experts Raise Concerns over EU Cyber Resilience Act’s Vulnerability Disclosure Requirements, CSO Online (Oct. 3, 2023) (open letter by experts and organisations including Google, the Electronic Frontier Foundation and Rapid7 objecting to the 24-hour reporting duty in art. 11 of the Commission’s proposal, now art. 14), https://www.csoonline.com/article/654175/cybersecurity-experts-raise-concerns-over-eu-cyber-resilience-acts-vulnerability-disclosure-requirements.html; see also Roumani, supra note 15, at 4 (ZDI gives vendors 120 days to release a patch).
74. Defense Export Control Law, 5766-2007 (Isr.) (export licensing administered by the Defense Export Controls Agency (DECA), Ministry of Defense).
75. Manohar Lal Sharma, supra note 49, ¶¶ 6, 42–44 (recording the July 2021 findings of an international consortium of journalistic organisations on the use of Pegasus, and that the petitions rested on news reports and on material such as the Citizen Lab reports).
76. Associated Press, Fearing Misuse, Israel Tightens Supervision of Cyber Exports, SecurityWeek (Dec. 6, 2021) (purchasing States must sign a declaration to use Israeli cyber products “for the investigation and prevention of terrorist acts and serious crimes only”), https://www.securityweek.com/fearing-misuse-israel-tightens-supervision-cyber-exports/.
77. Meir Orbach, Israel Defense Ministry Slashes Cyber Export List, Drops Saudi Arabia, UAE, Calcalist (CTech) (Nov. 25, 2021) (list of approved destination countries cut from 102 to 37), https://www.calcalistech.com/ctech/articles/0,7340,L-3923361,00.html.
78. Daniel Estrin, What to Know About the Spying Scandal Linked to Israeli Tech Firm NSO, NPR (Aug. 25, 2021) (quoting Israel’s Defense Ministry), https://www.npr.org/2021/08/25/1027397544/nso-group-pegasus-spyware-mobile-israel.
79. IT Act, supra note 8, § 75 (Act to apply to offences or contraventions committed outside India).
80. China Vulnerability Provisions, supra note 11, arts. 4, 7(2), 9(7), 15.
81. Ryan Naraine, Microsoft: China Flaw Disclosure Law Part of Zero-Day Exploit Surge, SecurityWeek (Nov. 7, 2022) (reporting Microsoft’s Digital Defense Report 2022: “This new regulation might enable elements in the Chinese government to stockpile reported vulnerabilities toward weaponizing them”), https://www.securityweek.com/microsoft-china-flaw-disclosure-law-part-zero-day-exploit-surge/.
82. Dellago, Woods & Simpson, supra note 2, at 3 (brokers may have legal entities in multiple countries to deal with export regulations).
83. Manohar Lal Sharma, supra note 49, ¶ 47 (free flow of information is an important step towards governmental transparency and openness).
84. Bilge & Dumitraş, supra note 4, at 839.
85. Manohar Lal Sharma, supra note 49.
