DeepSec 2025 Talk: Machine Learning Poisoning: How Attackers Can Manipulate AI Models for Malicious Purposes – Shahmeer Amir
The use of machine learning and artificial intelligence has been on the rise in various industries, including the field of cybersecurity. These technologies have shown great potential in detecting and mitigating cyber threats, but they also come with their own set of risks. One of the most significant risks is the threat of machine learning poisoning attacks.
Machine learning poisoning attacks involve an attacker manipulating the data or the learning algorithm used by an AI model to compromise its accuracy or functionality. This type of attack is particularly dangerous because it can go undetected for a long time, and it can be challenging to trace its origins. A successful poisoning attack can result in the AI model making incorrect decisions, which can lead to a security breach or data loss.
The session will cover practical steps that organizations can take to prevent machine learning poisoning attacks. These measures include data validation, monitoring the performance of AI models, and implementing adversarial training techniques. Attendees will learn to implement these measures and secure their systems from machine learning poisoning attacks. Attendees will understand how these attacks were carried out and the resulting lessons.
The presentation will also include case studies of high-profile machine learning poisoning attacks, highlighting the impact they had on the organizations targeted.
We asked Shahmeer a few more questions about his talk.
Please tell us the top facts about your talk.
- It will make you not trust AI Models
- You will know how easily AI models are poisoned
- Our reliance on AI models may cost us precious data/business
How did you come up with it? Was there something like an initial spark that set your mind on creating this talk?
I was always irritated of how people trusted AI models, how much they relied on them, I knew there was something something wrong had to dig in. Digging in I found out how easy it is to poison AI models
Why do you think this is an important topic?
World is quickly moving towards an AI-dependent culture, we need to know the risks before we can proceed
Is there something you want everybody to know – some good advice for our readers maybe?
Work to Learn, Not to Earn
A prediction for the future – what do you think will be the next innovations or future downfalls when it comes to your field of expertise / the topic of your talk in particular?
AI will not take over the world. It will result in losses of 100s of billions of dollars.
Shahmeer Amir stands as a globally recognized Entrepreneur and Ethical Hacker, awarded Entrepreneur of the year 2024 and also ranking as the third most accomplished bug hunter globally. His expertise has been instrumental in assisting over 400 Fortune companies, such as Facebook, Microsoft, Yahoo, and Twitter, in resolving critical security issues within their systems. Shahmeer’s entrepreneurial ventures in the technology realm have led to the establishment of multiple startups, with his current role involving the leadership of Speeqr, and minor involvement in Veiliux and Authiun
In his capacity as the CEO of Speeqr, Shahmeer’s flagship company is actively engaged in developing next-generation audio-video communication technologies. Additionally, he serves as the CEO of Veiliux, positioned as Asia’s inaugural mainstream Cyber Security startup with a presence in the Asia Pacific, UAE, and the UK. Authiun, another startup under Shahmeer’s leadership, offers a comprehensive password-less authentication solution tailored for the 21st century.
Furthermore, Shahmeer serves as the Cyber Security Advisor to the Ministry of Finance in the Government of Pakistan. His involvement spans various projects, including Deep Sea Tracking, Digital Transformation of Legislation, and the Digitization of Pakistani Cultural Content. As a testament to his influence in the tech industry, he holds a position on the Forbes Technology Council.
An engineer and certified Cyber Security professional with credentials from esteemed organizations like EC-Council, Mile2, SANS, among others, Shahmeer is currently exploring Blockchain technology for his doctorate. With three authored books, including “Bug Bounty Hunting Essentials,” and numerous research papers, he has solidified his standing as a thought leader in the field.
Shahmeer’s prominence extends to the speaking circuit, where he is a highly sought-after keynote speaker on topics such as Cyber Security, Blockchain, and various technologies. Having received invitations to over 80 conferences globally, including prestigious events like Blackhat, GiSec, FIC, AEC Alberta, and many other Global events, he continues to be a beacon of knowledge. Accepted into entrepreneurship programs at esteemed academic institutions, including Stanford, Shahmeer’s prowess is further demonstrated by his proficiency in coding in 25 languages and reading code in 35, establishing him as an expert across multiple technologies in his role as CTO of companies.